Rendered at 13:16:02 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
bob1029 16 hours ago [-]
One unit of HBM capacity consumes roughly the wafer capacity that could have produced three units of DDR5 capacity. HBM dies need to be larger than ordinary DRAM dies because of how the final packaging works.
> As discussed previously, the ramp of HBM production will constrain industry supply growth in non-HBM products. Industrywide, HBM3E consumes approximately three times the wafer supply as D5 to produce a given number of bits in the same technology node. With increased performance and packaging complexity, across the industry, we expect this trade ratio for HBM4 to be even higher than the trade ratio for HBM3E. We anticipate strong HBM demand due to AI, combined with increasing silicon intensity of the HBM roadmap, to contribute to tight supply conditions for DRAM across all end markets. As the memory industry is still recovering from the challenging environment in 2023, this tight supply environment will help drive the considerable improvements in profitability and ROI (return on investment) that are needed to enable the investments required to support future growth.
>As the memory industry is still recovering from the challenging environment in 2023, this tight supply environment will help drive the considerable improvements in profitability and ROI (return on investment) that are needed to enable the investments required to support future growth.
The problem is that they need actual chips tomorrow. I fear we are reaching the point in the semiconductor industry where, in order to sustain the revenue growth propping up their valuations, they're going to have to start selling future chips that cannot possibly be physically produced.
pmontra 4 hours ago [-]
But maybe no customer will be there to actually buy, because cash flows could crash their businesses. So options on chips? Call or put?
freeone3000 14 hours ago [-]
The alternative is they could charge more. They could. And people would pay it.
vannevar 14 hours ago [-]
The problem isn't the price. No matter how much you pay, you can't take delivery of a chip that doesn't exist. Now, you could raise prices to the point where you destroy demand. But that's a tricky window to maneuver through.
sidewndr46 2 hours ago [-]
Aren't the advertised prices just a number now? When OpenAI or some other big player comes to buy months worth of production capacity, they aren't doing it at a publicly advertised price.
HumblyTossed 13 hours ago [-]
If you can only make X number of widgets, you must increase the price until demand stabilizes at X.
cgio 4 hours ago [-]
The price will increase in secondary market.
Forgeties79 1 hours ago [-]
To some degree sure but how many people/businesses are tearing ram out of their existing hardware to resell?
docjay 12 hours ago [-]
Why must you increase the price?
cortesoft 11 hours ago [-]
The alternative is to have shortages.
If there is demand for N chips at $X price, but you only have N/2 chips, then half the people aren't going to be able to buy them. The people who really need the chips, and would be willing to pay a lot more than $X for them, will be competing with people who are only willing to pay $X for them. You will end up getting scalping and shortages and hoarding. Since people know that there are enough people willing to pay a higher price, everyone will try to buy them at $X, even if they don't need them at all. Just buy them at $X, and immediately sell it to one of those companies willing to pay a lot more.
The market is extremely inefficient... since the manufacturer isn't charging enough, you get way too many people trying to buy them from the manufacturer.
So you find the price where the demand is N/2 chips, and everyone who is willing to pay that much will get one, and there is no profit for scalpers so the only people who will buy them will be actual companies that need them.
franga2000 5 hours ago [-]
> The people who really need the chips, and would be willing to pay a lot more than $X for them
The people who really want the chips, and are able to pay a lot more than $X for them.
Please can y'all stop pretending that raising prices allocates resources by need and willingness to pay? With the level of spending power imbalance, the person who wants something and has too much money will ALWAYS squeeze out the people who actually need something and have less money. Always. No other factors are considered by the market.
wisty 4 hours ago [-]
The cure for high prices is high prices. Selling out creates scalpers and sometimes even a speculative bubble. High prices means producers invest in new fabs (though it may take time).
Forgeties79 1 hours ago [-]
>though it may take time
That’s what makes this so distinct so it probably shouldn’t just be some offhand parenthetical at the end. It’s incredibly expensive, it takes a long time, and it’s incredibly specialized. You can absolutely forget the idea of a “disruptive startup” entering that space too.
spockz 57 minutes ago [-]
Maybe not a traditional disruptive startup but isn’t CXMT from China stepping in to build all those fabs?
Previously investing in that space didn’t make a lot of sense. Now that there are a lot of consumers willing to buy something they can build and sell at a price between the previous low point and the current high point.
The only tricky thing is that the current shortages need to continue until the fans are well up and running.
Even if the supply side suddenly crashes due to the AI bubble bursting or someone inventing a way to only need 1/10 of RAM, I can imagine them keeping the prices relatively high.
inigyou 3 hours ago [-]
Can we stop abusing the technical economic definition of "shortage"? In economics a shortage means the price is too low for the demand quantity, but in everyday usage a shortage means the supply quantity is too low.
With the basic supply/demand curve, "shortage" and "discount" are synonyms and I know which word makes more sense to describe the scenario.
docjay 9 hours ago [-]
If demand outpaces supply then there’s a shortage no matter what you do.
I don’t believe scalpers are an issue in B2B; these aren’t concert tickets being sold to the general public.
It’s clear you grasp the economic theory as it’s taught, but my comment was meant for you to question it. If someone is hungry you can charge more for food. The market allows and encourages it. But don’t mistake that for “willingness” and don’t mistake raising the prices purely to get extra money from the exchange as some inevitable law of the universe. You’re welcome to love the concept, but don’t whitewash it.
crote 9 hours ago [-]
> If demand outpaces supply then there’s a shortage no matter what you do.
Demand is elastic. If the price of potatoes rises to $1000 / kg most people will switch to using pasta or rice for their dinner instead. There might be a shortage of potatoes at $1 / kg, but not at $1000 / kg.
If my 7-year-old computer were to die tomorrow I would've normally replace the entire thing. It has had a good life, the newer generations of hardware are a decent bunch faster, and it would be nice to get some additional features. With current RAM prices? No way I can afford to upgrade to DDR5, I'll have to get a replacement AM4 motherboard / CPU to fix it.
> I don’t believe scalpers are an issue in B2B
They are called "speculators". DRAM is a commodity and is traded no different from, say, potatoes. It's why there are companies like DRAMeXchange. DRAM module assemblers like Kingston, G.Skill, and Corsair will buy chips from whoever gives them the best deal. Similarly, anyone assembling hardware using DRAM will have stockpiled it when the AI boom became noticeable - with some almost certainly selling it off now that it has become too valuable to use for cheaper products.
swiftcoder 1 hours ago [-]
> Demand is elastic. If the price of potatoes rises to $1000 / kg most people will switch to using pasta or rice for their dinner instead. There might be a shortage of potatoes at $1 / kg, but not at $1000 / kg.
The year 1845 called, and they'd like a word
tsimionescu 3 hours ago [-]
> Demand is elastic. If the price of potatoes rises to $1000 / kg most people will switch to using pasta or rice for their dinner instead. There might be a shortage of potatoes at $1 / kg, but not at $1000 / kg.
If potatoes suddenly became $1000/kg, everyone would agree that there is a shortage of potatoes. While sure, people won't starve and will eat other things, it doesn't change the basic fact that there aren't enough potatoes.
The situation with RAM is even worse. If I'm trying to build and sell computing devices, I need RAM. Whether no one is willing to sell RAM at all, or whether they'll sell it only for 10x yesteryears's price, it doesn't make much difference to me, I'm still going out of business. Sure, that will then make demand technically go down, but not the shortage.
1 hours ago [-]
b112 3 hours ago [-]
This whole thread is a bad analogy. Take all potatoes out of the market, and:
1) Some people won't be able to afford alternatives. They will starve to death
2) There may not be enough of other foods, so even if people have money, people will starve to death.
There are years where corn or rice or oats are of poor yield. A couple of years of that, and canned food and stored food is gone. It doesn't take long for "we have no <core staple food>", and things become bad.
This isn't a 3 month or year long blip in RAM pricing. If nothing changes, no bubble bursts, we won't see RAM become plentiful until 2035 at least. And a tight supply until 2030.
That's not a year of no potatoes. That's a decade of a core crop, with loads of calories, gone. That's "people are going to starve somewhere" trouble.
(I hate this food analogy, because in the West we've not seen a drought cause people to starve for half a century, if not longer. People are too complacent.
To speak to that, we grow crops where we get best value for the money. And best results. Corn cannot be grown in much of Canada, it needs a LOT of sun to mature, where as wheat can with a shorter growing season. Rice needs lots of water, where as other crops are more tolerant. Potatoes do well in certain soils. If there are no potatoes, you cannot just grow anything in that location.
Take away a core crop, and you can just lose those calories. Especially as farmers won't switch on the first year of a lost crop.)
RAM is quite durable, so I can imagine buying new servers with new CPUs, but instead of DDR5, it has DDR4 RAM slots. A loss of performance to be sure, but a good way to renew your fleet, and reduce the impact of all of this. But such boards aren't really here I don't think, not in quantity.
jonhohle 8 hours ago [-]
I just did exactly that. I had an old i7 with 64GB of DDR4 that I bought when it was cheap. DDR5 was so expensive I just bought an AM4 and moved the RAM over.
My MacBook is showing its age, though. Hearing that 2027 wont be any better makes me hopeful all the vibe coders can ask for memory efficient implementations to stretch existing hardware a little further.
tjwebbnorfolk 9 hours ago [-]
When demand exceeds supply you have two options: make the people who value those goods the highest compete with each other to pay the natural market price for them, or have a lottery with price ceilings. Price ceilings create inefficiency, and they create black markets.
In a lottery, the goods are misallocated to a bunch of people who aren't getting the most value out of the goods, creating economic inefficiency. A bunch of GPUs would be sitting in warehouses waiting to be resold (either when the price goes up, or in the inevitable black market), or in my basement screwing around with them, instead of being deployed in a way that the most people can benefit from creating maximum economic value from their deployment.
Your gaming PC isn't as economically valuable as JPMorgan using AI for fraud detection, for example.
Moreover, if you are appointed god and force everyone to sell GPUs at one dollar just because you want cheap GPUs, then this is the last batch of GPUs that will ever be produced and you'll have a shortage until the end ot time.
repsilat 7 hours ago [-]
> you have two options
There are many other ways to allocate capacity. For example, you could give preferential access to customers who sign long-term purchase commitments, or some other favourable terms. These other methods might correlate well with value to the seller and might be easier for buyers to pay.
Other methods could be "allocate to customers that you want to maintain a good relationship with." I suspect HBM is not allocated either with pure auctions or by lottery.
tsimionescu 6 hours ago [-]
Not that I'm advocating for this with RAM, but another tried and true solution for necessary goods with limited supply is rationing. This is what got everyone through WW2, for example.
EliRivers 2 hours ago [-]
This all sounds very hypothetical economic universe. In the real world, rich people do buy expensive things and do nothing with them.
2 hours ago [-]
spoonjim 8 hours ago [-]
> The people who really need the chips, and would be willing to pay a lot more than $X for them, will be competing with people who are only willing to pay $X for them.
If there was an Ozempic shortage tomorrow, it's not the morbidly obese people who "really need" it who would be getting it, it would be the Hollywood A-listers who need to drop 4 pounds for their next movie, or the billionaire who's feeling a little tubby. Price is a good allocation system, but it's important to recognize that it's not a mechanism for attaining some kind of ethically "best" allocation.
inigyou 4 hours ago [-]
Because otherwise you will be outcompeted. In other words capitalism will fire you for doing a bad job.
LoganDark 11 hours ago [-]
Because if you take orders otherwise you won't be able to fulfill them.
somat 11 hours ago [-]
It is the other way around, When people want something they are willing to pay more for it. The manufacturer wants to sell to the person that will pay the most. While the person wants to buy from the manufacturer that charges the least. As such when the manufacturing capacity is below demand the price goes up until the item is worth what people will pay for it(at large, statistically, it works like an auction). The promise of capitalism is that as the price goes up it should incentivize manufacturing capacity to go up and as the capacity meets demand and the manufactures start to fight each other to be the one to sell the item the price goes down.
Sometimes this works and sometimes it does not.
docjay 9 hours ago [-]
Thank you for taking the time and effort to provide a clear explanation. I’m familiar with supply and demand though. My emphasis was on “must” as a lazy protest against greed being considered a mandatory part of capitalism. Prices don’t have to go up, someone just wants more money.
If you happily sell apples for $1 and see a hungry person walking towards you, must you raise the price?
I think that basic thought gets lost sometimes when people talk about shortages causing the price to increase. The shortage didn’t cause anything, some executive decided they want more money. That’s all. There’s nothing inherent in the system that requires it. Whether that’s OK or not is up to the reader, I just think people lose sight of the reality and talk about it like it’s gravity, rather than simple decision making.
ShinyLeftPad 3 hours ago [-]
> If you happily sell apples for $1 and see a hungry person walking towards you, must you raise the price?
if you happily sell apples for $1 and now there are VC-funded companies who need a ton of apples to produce some dubious crap and can pay you $5 per. And you are the only company selling apples.
what you should do is keep selling $1 apples to retail customers and charge corporations as much as they accept. you invest it into more apple trees and maybe subsidize the hungry people even
1718627440 1 hours ago [-]
The issue with that solution is that in order to sell to retail customers, they sell to other large corporations, which they think are going to sell to retail customers. Guess what happens when they find out, you are charging other corporation higher prices.
HWR_14 13 hours ago [-]
It could be tricky. Or, since supply (in the short term) is pretty well established, they could just auction off chips.
boredatoms 14 hours ago [-]
For consumers they’re already destroying demand
daishi55 11 hours ago [-]
I don’t understand how there being essentially unlimited demand for their products that far exceeds supply and is driving up prices accordingly is somehow a bad sign for the industry?
> Now, you could raise prices to the point where you destroy demand.
You know supply and demand is like a curve right, you can find an optimal equilibrium? It’s not a cliff that you can fall off.
vannevar 7 hours ago [-]
"Demand destruction" doesn't literally mean all demand is destroyed; it refers to the demand curve you reference. But as a sibling commenter notes, real business is rarely as clean as an econ textbook. In the real world, a supplier can contract to supply more units than they can actually produce. Their customers rely on the representation to make other related deals. And of course, the supplier can book the contracted revenue, causing investors to rely on the forward sales. They can't raise the price at that point, it's fixed in the contract. So if it turns out they physically can't deliver when the time comes, some number of the deals have to be blown up, causing related deals to blow up, etc., etc. That is the risk I was referring to.
daishi55 3 minutes ago [-]
Well, let’s see if overwhelming demand for their products turns out to be good or bad for the semiconductor industry. I think the answer is pretty obvious.
noduerme 4 hours ago [-]
Or: Supply and demand oscillate in a delayed way based on each other. A quick increase or decrease in one side takes time to ripple, and by the time it does, the organization may have reversed course.
An airline overbooks flights. Lots of people get their flights cancelled. Takes six months for the fallout to settle where everyone who got burned booked their future flights on a different airline. 3 months out the airline has to cut the number of flights and cut prices in the face of falling revenue to reclaim market share. As soon as they do that, they're flooded with too many bookings. So they take the bookings and overbook flights again, but it takes time to bring the new flights online.
It's not a 1:1 situation. Each time you miss the market you wobble a little further until all the inefficiencies of bad predictions eat you up.
daishi55 2 minutes ago [-]
None of this suggests to me that enormous demand for semiconductors could be bad for the semiconductor industry.
inigyou 3 hours ago [-]
It's good for the memory chip industry. It's bad for almost every other one.
doubletwoyou 11 hours ago [-]
It’s ideally modeled as a curve, but things are never that simple in real life.
Jgrubb 11 hours ago [-]
I guess we'll see about that.
fn-mote 13 hours ago [-]
> The problem isn't the price. No matter how much you pay, you can't take delivery of a chip that doesn't exist
I’m confused because this reads like a denial of basic economics. If the price is high enough, the chip will be produced for you.
Do you mean because of the production lead time, higher prices won’t result in increased production? Commodities like corn have been managing this for a long time… what’s special about chips?
What is your actual argument?
crote 9 hours ago [-]
Nontrivial markets don't behave like Economics 101 textbook examples.
Samsung, SK Hynix, and Micron have a combined market share of 90% - with most of the rest being a very new-to-the-market CXMT. It is a cutthroat market which behaves like a stereotypical "pork cycle". Semiconductor fabs cost billions to build and take years to complete, so you better be damn sure you have buyers before you start constructing one. You and your competition overestimated the demand? You have to pay back the construction cost, so you're now in a race to the bottom and one of you is going bankrupt.
Ever wondered where Intel came from? They started out as a DRAM manufacturer, which dominated their revenue well after the introduction of their first microprocessors. But in the early 1980s the glut of supply from new Japanese manufacturers made it so unprofitable that they had to ditch the memory market altogether. The stories of Texas Instruments and Motorola aren't much different. And that's not even mentioning the likes of Mostek, which once held a 85% market share and was dead less than 5 years later! Oh, and those Japanese manufacturers? All gone, pivoted like Intel or died like Mostek.
So no, the three remaining DRAM manufacturers aren't going behave like headless chickens and start ordering new fabs just because there's a bubble causing a temporary demand peak. Unless those AI companies are going to pay in advance, in cash, for an entire fab, they'll just have to wait and deal with the price increase.
DannyBee 13 hours ago [-]
Corn is heavily subsidized,also has (in the US) federally prvided insurance programs against plunging prices, has paid uselessness(ethanol). Despite all of this corn has caused massive agriculture bankruptcy/takeover when it has failed anyway.
So it's more "what's special about corn". It is also fairly hilarious to claim the parent is denying basic economics and then bring up corn as an example of having successfully managed economics. If the scales were not being thumbed, and "basic economics" were in play, corn would be in very very bad shape.
In the case of DRAM, there is an incredibly long history of these gloom/glut cycles, and they have stayed roughly the same timeframes (~3 years) since the 1990's.
Almost all the ones who have survived this long are either in the same kind of boat as corn - protected in various forms from the downside - or don't increase production and get caught out until they are absoultely forced.
The very temporarily increased profit is not worth going bankrupt for - they make more money long term by being very cautious and know this.
There are a near infinite number of economic studies you could look at (and several sibling comments cite some) - DRAM manufactuers don't chase the price and probably couldn't anymore if they want to.
None of this denies basic economic theory, of course, since economic theory is not exactly "rigorous", even to the degree it could be (IE even the parts that are pure analysis of data rarely reproduce!).
SR2Z 11 hours ago [-]
I don't think that anything you said is wrong but I also don't think that memory consumption is going back to the old baseline... well, ever.
Memory is just too useful now that you can use it to drive cars and write code.
crote 9 hours ago [-]
People are happy to play with AI when the tech companies are burning hundreds of billions of dollars to subsidize it. It remains to be seen who is actually willing to pay for it at the prices required to recoup those insane investments.
Which directly leads to the next big development: all the big players are investing in silicon with "baked-in" models, like [0,1]. Turns out you don't need an expensive general-purpose GPU with heaps of RAM to contain a model when you can make a custom ASIC around one specific model! Why spend a fortune on DRAM / HBM when all you need is some finetuning parameters which are easily stored in on-die SRAM?
> It remains to be seen who is actually willing to pay for it at the prices required to recoup those insane investments.
Everyone. Open models are going to keep prices down. A lot of current models are more than usable. Self-hosting would have been an option if hardware prices get back to sane values.
The insane investments have to do with insane over-valuations, VCs involved and hence media spam on it. Chinese labs for example make do with 1% of the valuation and 1% of the resources.
pmontra 3 hours ago [-]
Prices must raise a lot to make self hosting mainstream or at least fairly popular.
Yesterday a coworker posted on a customer's Slack the specs of a box he is planning to buy to run local models. It's about 5k Euro. I am paying 18 Euro per month for Claude Pro and I'm going through a migration (almost a total rewrite) of a web app from Vue 2 / Vuetify 2 / Vuex to Vue 3 / Vuetify 4 / Pinia. I never hit the 6 hours limit. I could consider running an equivalent model on a 500 Euro machine (about 2 years of Claude Pro) but 5k is 20 years and that box will be obsolete or will have failed beyond repair (no spares) much earlier than that.
nemothekid 10 hours ago [-]
The problem is that is speculation and the market barriers are currently too high for losing business to be a threat. The two outcomes are (1) you overbuild, you end being wrong, you go bankrupt and lose everything or (2) you are right, but since you didn't overbuild, you lost out on some revenue, but demand still exists 3 years later and you didn't lose everything.
In any other business choosing (2) would mean someone else swoops in and steals all your business. It doesn't look like this is at all possible for memory fabs.
DannyBee 36 minutes ago [-]
"In any other business choosing (2) would mean someone else swoops in and steals all your business. It doesn't look like this is at all possible for memory fabs."
FWIW - Lots of people seem to believe the "in any other business", but it turns out there are tons and tons of places it is equally untrue, even where there is near zero time or cost barrier to entry.
It makes sense, CXMT takes as much profit as possible, then use that money to build more capacity, ultimately becoming one of the biggest competitor (only?) to the cartel.
crote 9 hours ago [-]
As of Q1 2026: 90% market share for the Big Three, 8% CXMT, 2% everyone else. And CXMT was only 3% Q1 2025.
1718627440 1 hours ago [-]
> Memory is just too useful now that you can use it to drive cars and write code.
You can use the state of the art from 20 years ago for that.
1718627440 1 hours ago [-]
> I’m confused because this reads like a denial of basic economics. If the price is high enough, the chip will be produced for you.
That's a classic stereotype of what an economic "expert" is going to say, while completely ignoring reality. Just because someone is willing to pay for it, does not mean, that goods just materialize. Sure, a lot of the time someone finds a way, but that is still limited by physical constraints: lack of expertise, lack of needed resources or as in this case simply lack of time.
JoeAltmaier 13 hours ago [-]
It takes a billion dollars and years to produce a new chip fab line. So no, it cannot just be produced for you on a dime.
XorNot 12 hours ago [-]
Also supply contracts exist which is what started all this: OpenAI signed huge orders for most of the market all at once. No one had any chance to properly include that information in their pricing or bidding.
p1necone 12 hours ago [-]
The capital investment needed to bring new chip fabs online and to staff them is likely orders of magnitude higher than that needed to buy land to grow corn on. And then the ratio of investment to sell price on that land + infrastructure is probably significantly worse for chip fabs that potentially aren't needed to satisfy demand anymore a few years from now.
"If the price is high enough" is of course technically true, but the scale of what high means in this context is important.
3 hours ago [-]
vannevar 7 hours ago [-]
>Commodities like corn have been managing this for a long time… what’s special about chips?
Chips aren't grown, for starters. They require very expensive facilities that take a long time to plan, build and start up. And ag firms are not seeing their stock price skyrocket based on the assumption that they will show exponential revenue growth every quarter, so there's no incentive for them to overbook their capacity and hope for the best. Maybe if the population was doubling every six months, you'd see something similar in corn. Prices in the real world are not perfectly elastic, and big deals are often locked in long before actual production.
inigyou 3 hours ago [-]
How much would you produce a RAM chip for me for? Not buy - produce.
It's a frictionless market in a vacuum right? No barriers to entry. You could make one right now if you wanted.
juiceland 13 hours ago [-]
[flagged]
toss1 11 hours ago [-]
Paying more would not help you when there is no more supply. It is a zero sum game.
Maybe you could pay someone else who is less desperate to part with some, but that does not increase supply.
Nine women cannot produce a baby in a month, even though you could average about one per month if you wait about nine months.
HumblyTossed 13 hours ago [-]
This is what consumers are always told. Prices go up as demand goes up or scarcity increases.
sandworm101 12 hours ago [-]
But once you get past econ101, you are taught how markets react to percieved bubbles, how those with near monopolies would rather pocket a windfall now than risk investing in future expansions.
dismalaf 11 hours ago [-]
Well, look at AI. It sure looks like a bubble, why wouldn't every fab just profiteer now?
scrtm 11 hours ago [-]
micron is up to like 85% gross margin this year
re-thc 6 hours ago [-]
> The problem is that they need actual chips tomorrow.
They need it today. Best if it was NOW!
adastra22 8 hours ago [-]
We saw this play out in Bitcoin mining years ago.
tverbeure 14 hours ago [-]
> HBM3E consumes approximately three times the wafer supply as D5 to produce a given number of bits in the same technology node
That's really interesting, and I wonder why? I believe HBM has redundant ECC bits by default, which would add a few %, but other than that, is it just that the yield is much lower due to die stacking? Of course, this is a 2024 document so things may have changed a bit since.
benjiro29 13 hours ago [-]
HBM requires stacking the chips. So they need to shave the layers, glue, stack more, shave again. They also require a substrate what is even more wafers. The issue is that a error in the stack means a lot of losses.
In order to get high bandwidth, you want memory as close as possible to the GPU. The more trace lane length = signal loss, bandwidth loss.. HBM is compact, and so you can stack 24GB modules, 8 around a GPU die.
If you tried to do that with normal memory, you need like 64 modules. So a a TON of traces more that all need to be equal length, and because so many = far away from the GPU = less bandwidth.
The issue is like stated above, its a process that waste a ton of wafers. Wafers that can make easily 3x more normal memory.
Intel with "Crescent Island" is trying to make a 160GB card using LPDDR5x memory but the bandwidth is only ~700GB/s.
wtallis 11 hours ago [-]
None of what you said actually explains how HBM could require 3x the wafers.
The base interface die has eight or 12 memory dies stacked on top of it. The wafer cost of that interface die is therefore small.
If the process of die thinning and TSV stacking reduced yields by a factor of three, nobody would consider HBM mature enough to put into production, and especially not mature enough to be increasing stack height from one generation to the next.
Trace length has approximately nothing to do with die size. If anything, designing for shorter traces means you can get away with smaller PHYs at either end.
What might go some ways toward explaining such a huge difference in die size is that the TSVs themselves take up significant die area and must be fairly numerous to carry both a large number of signal wires and all the power and ground required by the stack. But it's wildly implausible that the die area consumed by the TSVs would be significantly larger than the die area consumed by the memory arrays themselves, or that anyone would build a memory die where the memory array was not a large majority of the total die area.
If there's any truth to that ~3x higher wafer requirement for the same number of bits as compared to DDR5, it must be a combination of several factors and probably includes something non-obvious and dubious, like counting all the area of the passive interposers that go between HBM stacks and GPUs (those interposers aren't competing for the same fab space).
MayeulC 3 hours ago [-]
Yeah, I'm dubious of the "3x wafers" claim, but I heard "3x wafer capacity", which sounds more logical. More steps per wafer (for TSVs) -> more time spent in the fan per wafer & machines busy with HBM wafer -> smaller wafer / chip output.
10 hours ago [-]
throwatdem12311 13 hours ago [-]
My own PC died and now I’m locked out of my entire Steam library for the foreseeable future. A $2000 PC is literally a downgrade from what I bought 10 freaking years ago.
What the freaking hell.
damontal 8 minutes ago [-]
My feeds are full of ads by gaming pc makers now renting out gaming pcs for a monthly fee. NZXT calls it a lease with a 2 year refresh.
the__alchemist 58 minutes ago [-]
I hear you. I'm the fool who bought an RTX4080 because "$1600 is too much for a GPU". (RTX4090; shipped to your door from the nvidia website). I don't think there's any reasonable upgrade path right nor or in the near future which would have been better than just buying that.
stephenitis 13 hours ago [-]
What part of your PC died?
throwatdem12311 12 hours ago [-]
Motherboard. New motherboard…so new cpu socket (mine was Broadwell)…and new ram needed.
GPU is mostly fine and can probably reuse that but it’s Vega so not the greatest I was due for an upgrade but the prices of everything is ridiculous.
The thought of pc-of-Theseus’ing my busted ass PC is just sad.
Thank god I bought a PS5 Pro before Sony started jacking up the prices.
I remember when buying a $300 GPU was extravagant and gave you state of the art graphics.
toaste_ 10 hours ago [-]
So, you have DDR4?
Pick any B550 with a VRM heatsink for $90, a 5600 for $190, and you are back in business.
Get a 5070 or 9070 before GDDR supply is completely gone if you want an actual upgrade.
As for GPU price, that Vega 64 was $500 in 2018. Even a midrange GPU hasn't been $300 since the Radeon RX 480 launched a decade ago.
SuperNinKenDo 10 hours ago [-]
This is essentially what I did and it worked a treat. Get in while the getting's still good. Having a modern but DDR4 compatible system is not obly a great way to get up and running now, but will help see you through this crisis long term.
jquery 6 hours ago [-]
Reading this thread I am so frickin glad I built a monster DDR4 workstation back in 2021. It would cost me more to build out that computer today (dual 3090s, 512GB ECC, 3995wx). Ram was so cheap back then.
clircle 2 hours ago [-]
Very neat. What's the power draw on this machine?
pdimitar 12 hours ago [-]
Why don't you just get a MB only for now? What mandates an entire new batch of hardware?
throwatdem12311 11 hours ago [-]
I’ve been pricing out my various options but the notion of replacing a 10 year old motherboard with another 10 year old motherboard is kinda lousy.
The PC I replaced when I bought this one was such a huge improvement.
My options basically boils down to fixing trash that’s falling apart or a sidegrade at best for the same price as what I bought a beast of a PC for 10 freaking years ago. Depressing.
Krssst 8 hours ago [-]
Buying a cheap secondhand motherboard is still less lousy than losing access to a game collection. I did that with a 13 years old computer and it cost 50€.
I agree that the current situation sucks, but I don't think one can say it locks someone on your situation out of your game collection.
(edit: misremembered, $2000 is not the price of the game collection)
starkrights 11 hours ago [-]
10 year old is still DDR4, and ddr4 is still plenty fine for gaming workloads. AM4 still has some solid options out there, the 5800x3d has 3d vcache and would be an insane upgrade over any… quad core? 6th gen intel? Not sure what the AMD at the time was.
14th gen intel supports both ddr4/5 if you wanted the ram upgrade potential, but it… well, I remember it being pretty sketchy perf/degradation wise for a while, so maybe don’t go with them.
I understand the sucky feeling that you’re sidegrading instead of investing in a new beastly rig for a reasonable price- believe me I’ve been in the same boat since all this GPU bullshit started. I got lucky choosing to upgrade to ddr5 years ago, but I elected not to get a GPU out of a lack of clairvoyance :(.
Still though- I upgraded from an i7 6700k in 2023. The 5800x3d released a year prior to the CPU I picked up. Granted I went to zen 5, but holy shit I didn’t not realize how fucking old that 6700k was haha. The 3dvcache on AMD CPUs is a crazy, crazy boost- in addition to the fact that modern CPUs are already bounds and leagues ahead of 2016, and I would go to bet that even if it’s technically a sidegrade, you probably wouldn’t feel too horrible about the money spent. (I really want to say that you’d feel pretty satisfied with it it, but I can’t discount that I also jumped up to a higher tier zen 5 cpu, and don’t have explicit experience with the 5800x or other AM4 stuff. ymmv!)
throwatdem12311 30 minutes ago [-]
Thanks for the advice gone look into it more this weekend
calgoo 5 hours ago [-]
Same, which is why im looking at some older epyc CPUs + ddr4 memory. the only issue being that i wont be able to use my current DDR4 ram in that new system but still... just hunting those ebay listings for a good combo from someone trusted (preferably local)
bdcravens 8 hours ago [-]
If it hadn't died, you'd still be running it.
pdimitar 11 hours ago [-]
Oh I agree, I am just saying that if you really did not want to spend, there are obviously options. I am thinking of promoting my super old and long-time-not-used PC with an i7-3770 and 32GB DDR3 RAM to a server and I bought a ~$45 motherboard for it some months ago, now I have to bother finding it a good NAS case and buy at least two HDDs (the real cost here).
Point is, quite a lot of hardware can still do a lot. I have a Ryzen 5500U-based laptop that still outperforms the PCs of most people that I know (non-tech people).
Though in my case I feel like I have to get a loan to buy a workstation + an actual gaming PC for my wife. That all is beginning to look like at least $15k... yikes. But I am not sure I can wait until 2028 to see if things _start_ normalizing... Depressing indeed.
alasano 1 hours ago [-]
My Truenas setup runs on a i7-3770 with 16GB DDR3, still good tbh. Have a whole bunch of self hosted apps running on it exposed through tailscale.
milch 8 hours ago [-]
I've been in the same boat since it seems something about the memory controller on my gaming PC's MB is broken, but price wise it feels like it just doesn't make sense. I could get something from eBay relatively cheap but then I'd probably spend an hour disassembling and reassembling just to maybe learn that the eBay unit has some equivalent issue or is slightly broken in some other way. I could old new stock for outrageous prices. Right now I'd rather just wait until either an upgrade would be REALLY worth it or at least prices come down.
inigyou 3 hours ago [-]
Last time I replaced an old MB it was only $60. What prices are you seeing?
It wasn't the same model or vendor MB but it was a compatible one. Same chipset.
rpunkfu 7 hours ago [-]
It sounds like whining for the sake of it really.
ed_mercer 12 hours ago [-]
can't you just buy a used same class motherboard from a marketplace?
cmrdporcupine 1 hours ago [-]
Similar thing happened here with my kid's 10 year old gaming PC. Motherboard dead. Had an existing 1070 and 16GB of DDR4. New DDR5 wasn't going to happen. New GPU wasn't going to happen. So I just got a Ryzen5 and an AM5 motherboard and re-used everything else. It runs just fine. It's fast enough.
Same I just built a dual Broadwell Xeon E5 server / NAS. Why? Because someone local was selling a cheap Chinese X99 motherboard, two Xeons, and 64GB of DDR4 RAM for $150 CAD. The deal was too good. So I built a machine around it. (Though what I should have done was shove it into an old Dell T630/730, and saved myself some hassle.)
Now is the time of re-using old tech.
inigyou 3 hours ago [-]
Buy an old motherboard compatible with your existing CPU. Reuse the CPU and RAM.
HaloZero 11 hours ago [-]
How does the steam machine compare to building your own?
kllrnohj 10 hours ago [-]
Building your own is cheaper, although not by a ton. Valve isn't taking a loss on them.
milch 8 hours ago [-]
Yeah, just barely, and only if you don't value warranty or build time. I'm also not super sure you could get something together that is somewhat compact like the steam machine and still be cheaper.
I've built two PCs in the last 5 years and both have had parts that were broken from the factory... There is something to be said about being able to hand off troubleshooting and replacement. I had ASRock pointing fingers at everything else to the point where I had swapped out CPU, PSU, and RAM, until they relented and replaced the motherboard (the replacement which is broken again now 5 years later by the way...)
Chrontius 3 hours ago [-]
It has a lot of benefits that aren't easily priced in.
bahmboo 12 hours ago [-]
There's always the BC 250 route. Takes some work but seems cool.
Mistletoe 10 hours ago [-]
Use the money to short the market because it’s probably a top signal of some sort.
kotaKat 3 hours ago [-]
I was in a shittier financial situation the past several years, finally get things situated to finally be able to buy a miniPC or one of those cool mini-laptops... then the Trump tariffs decimated the entire market.
That finally starts to ease pressure... then the AI RAM boom happens and now I can't even buy memory for anything I've got and I'm trapped in 8GB baseline RAM hell.
Every plan I had to update my network, my home, all out the window. I haven't seen my Steam library in at least a year or two now.
noduerme 4 hours ago [-]
So LLM profiteers have bought up all the consumer RAM for the next several years in order to deliver more of a service we're already oversaturated with that doesn't turn a profit..? I guess by then there'll be no consumers left with jobs to afford a PlayStation 6, so no problem. Except how will the AI trainers find revenue? If maximum potential is reached by laying off 90% of workers at companies that sell stuff... help me with the logic here... who's gonna buy the stuff those companies sell?
linsomniac 20 minutes ago [-]
>doesn't turn a profit.
This isn't the condemnation of AI that you think it is.
When you are building an industry, you don't want to be in a "cash cow" phase. Because the person who is willing to reinvest profits (and further: seek and sink investments) into growth and innovation is going to eat your lunch.
~8 years ago my BIL condemned Tesla with "they aren't profitable" and I explained this to him. Now they're a trillion dollar company.
Arodex 7 minutes ago [-]
You are basically describing predatory pricing or dumping, where the goal is not to be the best but to starve your competitors to build a monopoly and gouge customers later.
Uber tried that and failed. Now they cost as much as normal taxis if not more. Airbnb tried that and now cost as much as hotels if not more.
And tesla is a trillion dollar company on paper only. It can vanish into thin air in a day.
suncore 4 hours ago [-]
Not sure there is much actual profit in LLMs. It will be interesting to see promise meet reality, indeed.
Cthulhu_ 3 hours ago [-]
It's once again a market where there's morbillions of money being invested but no money is being made yet. The next few years will show whether AI becomes established in e.g. large enterprises who use it as part of their daily business. But there's a lot of competition and wishes to cut costs as well so profit margins will dwindle.
I also think most hardware expenditures done today will lose their value quickly; as we saw with crypto, hardware manufacturers will make specialized hardware so the cost/performance ratio will go down by at least 90% within a few years, possibly more.
noduerme 4 hours ago [-]
I mean, it's obviously that they think they're going way beyond LLMs to get to something that is profitable, like some ultimate goldmine of intelligence. But if they miss that target by a hair or two, can we buy some of that RAM? I'd like my next MacBook not to cost $10,000.
SecretDreams 2 hours ago [-]
I'm going to be very pumped when this comes crashing down.
It's going to look a bit like oligarchs subsidizing consumer tech as they try to offload more ram than God could own.
SecretDreams 2 hours ago [-]
The profit is in the stock price rising, not the actual revenue. They need stories to keep the stocks going up.
visarga 4 hours ago [-]
Medium term it could lower the price of memory and compute.
tiotempestade 4 hours ago [-]
Hoping all these DCs and AI whatnots turn out to be a flop, so that we’ll get plenty ddr5 at sale price :D
hinata08 2 hours ago [-]
even if they're a flop, you'll be told they're too big to fail.
And now, you have to bail them out and buy their HBM instead of paying for your RAM
I hope that China makes up for the capacity that RAM producers are not willing to produce, so that the latter are sent back to the past (with no bailout nor protectionism)
catlikesshrimp 2 hours ago [-]
You hope you and I will depend more on China? That won't fix any unbalanced market, it will move you another step towards a vertical structure..
esskay 4 hours ago [-]
IIRC this wont be possible as a lot of them aren't being packaged on traditional DDR memory boards. They're being put on things like the MI350X.
I'm sure there will be a ton of standard DDR5 sticks, but not so much from AI-only datacenters.
embedding-shape 3 hours ago [-]
> this wont be possible as a lot of them aren't being packaged on traditional DDR memory boards. They're being put on things like the MI350X
Even better, looking forward to having hardware like that at home for "cheap" in the future, compared to what it costs right now :)
inigyou 3 hours ago [-]
China has a cottage industry around taking apart circuit boards and reusing the chips in unexpected ways. That's how you get desktop MBs with soldered laptop CPUs. HBM and DDR aren't compatible as far as I'm aware, but they might be able to do something. Maybe they'll invent a converter chip.
cryo32 4 hours ago [-]
Expecting a pivot back to transitional parts if it goes bang.
pu_pe 1 days ago [-]
This will have general inflationary consequences for consumer products (phones, consoles, laptops, etc.). On top of current uncertainties regarding oil and fertilizers, I think 2% inflation in the US and Europe would be a very optimistic target.
Aurornis 17 hours ago [-]
Yes, but the impact will be small.
The average consumer isn't purchasing a lot of products with a lot of RAM every year.
Their 8GB of RAM phones will go up in price a little bit, but people aren't buying phones every year or even every other year.
So if the price of the 8GB of LPDDR went from $40 to $160 and it's all passed on to the consumer buying a new phone every 4 years, that's an extra $30/year in spending.
Most adults I know now use their phone for everything, so they're not buying new computers and laptops. They can wait 3 years for the market to settle before upgrading those, too.
Even I'm a heavy buyer of tech products and RAM, and I would bet you that my family's annual food bill fluctuates by more than what I've had to pay for RAM prices growing.
ambicapter 13 hours ago [-]
You don't have to directly buy RAM to be affected. It's just like oil or electricity...if RAM was part of the economic process that you participated in, then that price increase will be passed down to you. If every datacenter in the US suddenly had to pay double for electricity, how would that affect every operation that is operating in that datacenter? Ok now it's not electricity, it's the RAM.
rapjr9 14 hours ago [-]
The headline says "sold out". That doesn't mean prices will go up, that means the memory will not be available at all for many companies and the products they make will not be available to consumers.
Dylan16807 12 hours ago [-]
The companies making phones will be able to get new memory chips.
Many smaller companies will have issues getting enough, but even they could buy something with memory and rip it out. When you're willing to pay a hefty multiple, you can outbid other people. The shortages are not nearly bad enough to escape standard supply and demand.
raincole 11 hours ago [-]
It means prices will go up and that's it. If you read "sold out" and immediately believe "no company will have memory in 2027 besides AI companies" then I think you have to reconnect your thoughts to the reality.
pixl97 15 hours ago [-]
You forgot to tack on the rise in prices of SSDs too, so that's $50 per year at least.
>Most adults I know now use their phone for everything
You mean the cloud for everything. A lot of phone tasks share their computational and storage workloads off in the invisible ether that has actual computers with real costs behind them. Amazon has no problem with bumping up their costs in order to pay for their fleet of servers. I mean, what business doesn't use the cloud these days.
smt88 15 hours ago [-]
Almost everything we consume is downstream of RAM pricing. I don’t buy a phone every year, but corporate workforces, streaming servers, cars, factories, etc. will all cost more next year.
emacdona 13 hours ago [-]
Oh, I hope you are right about cars. If one of the outcomes of the memory crunch is that manufacturers start making “dumb cars” again, it would almost make it all worthwhile.
brewdad 9 hours ago [-]
Cars get developed on a cycle that spans a decade in many cases. Four or five years is pretty much the minimum. It could impact development cycles going forward but it will take a good while for that to be reflected in the marketplace.
In the shorter term, you’ll probably see more “base” models on offer with some of the most intensive features disabled. There’s still going to be all of the telemetry stuff that makes them money through third parties though.
lowbloodsugar 13 hours ago [-]
Remember when car companies couldn’t get the chips they needed to make cars? And so now there’s not enough used cars. And so car companies charge way more the new ones and just stopped making cheap cars altogether?
ragall 15 hours ago [-]
At the low end, where the vast majority of people are, even in the US, an extra $120 will double the price.
tonyedgecombe 1 days ago [-]
2% was already optimistic once we decided we were buying too much from China.
UncleOxidant 16 hours ago [-]
And once we also decided that we needed to attack Iran.
inigyou 18 hours ago [-]
Literally put a 50% tax on everything and then said it'll only raise prices 2%
Dylan16807 12 hours ago [-]
Imports from China to the US are around 1% of US GDP.
Total imports are maybe 14%? But not as highly tariffed.
tonyedgecombe 6 hours ago [-]
Yes, they are very cheap aren’t they.
MattDamonSpace 16 hours ago [-]
Why are “demand and supply driven price increases” considered inflationary?
Isn’t it just… prices?
pkulak 15 hours ago [-]
Prices rising is inflation.
tripleee 6 hours ago [-]
from what I understood it, at least originally, was used to mean monetary inflation
No, that's copium from libertarians. I don't have an issue from them also using the word inflation to describe something getting bigger, but it's different from prices going up, which is the definition of inflation.
odo1242 15 hours ago [-]
Any price increase can be considered inflationary. They usually don’t come down by all that much
Dylan16807 12 hours ago [-]
It's inflationary when the effect is big enough to have a meaningful impact on the price of the average good.
applfanboysbgon 15 hours ago [-]
It's inflationary because this demand is backed by Monopoly money. OpenAI is promising to buy half the global supply of RAM with money they don't actually have, pricing out people who would like to make purchases with real money and not $1 trillion in debt.
dgellow 1 days ago [-]
Yep. And not in the future, it is here right now. And poor countries will as always be impacted way more than the US or Europe. The AI industry is the very direct cause and should be shamed for handling their rollout in such a destructing way. They deliberately created and grew the feeling of urgency for their own benefit
honkostani 24 hours ago [-]
[dead]
apple4ever 22 hours ago [-]
This is why I'm so hesitant to use AI. There are other reasons (for example: the journey is the fun part, not the destination; and I want to learn not have something done for me), but the pressure AI puts on memory and storage is too much for me. I do use it occasionally, mostly to keep tabs on it and for small problem solving (how do I do this one piece of a query in MySQL, since I'm not a MySQL DBA), but that's it for now, until this memory/storage issue is fixed.
rickydroll 22 hours ago [-]
Unfortunately, I can't not use AI because it provides the best speech recognition I've ever used and, going along with hand problems, it now allows me to write code after a three-year absence from that skill. I do miss it. I'd never do it again professionally in the large, but I have missed writing code a lot.
My solution is to see what four or five-year-old equipment I can buy that will let me run local LLMs. I may only get six or seven tokens per second out of an i7, but it's a start. And best of all, I can turn the machine off when I'm not using it.
IMO, Migrating to small-scale local LLMs would be a significant improvement over using data centers.
Aurornis 14 hours ago [-]
> IMO, Migrating to small-scale local LLMs would be a significant improvement over using data centers.
LLM serving is most efficient when you batch a lot of parallel requests together. Data center solutions also have the advantage of collecting queries from around the globe, so the hardware can be utilized around the clock.
Having everyone serve their own local LLMs would produce a lot more memory demand. Not less. The same memory would be idle most of the time, and when it was used it would be used for 1 person instead of a batch of requests.
There are other reasons to run local LLMs, but solving hardware demand problems is not one of them.
theLiminator 14 hours ago [-]
Yeah, local LLMs are probably an order of magnitude less efficient at a fixed level of "intelligence" if not more.
tjwebbnorfolk 14 hours ago [-]
Inefficient per watt, yes, but local inference capacity is greatly underutilized in aggregate. If a model can run on a machine that already exists, that's a bunch of additional chips that don't need to be built.
sfRattan 15 hours ago [-]
> Migrating to small-scale local LLMs would be a significant improvement over using data centers.
This shift is probably inevitable, but it will vary significantly by region depending on prices for electricity. Look at, for example, the difference in fundamental homelab build recommendations between Germans and just about anyone else. Electricity prices in Germany are so high that even a now expensive Raspberry Pi or other ARM board is often preferred over Intel/AMD builds due to low power draw (especially low idle power draw), an effect that adds up for a machine running all the time over years.
With local LLMs and the GPUs to run it, especially if you want a model available to you all the time and can remote into your local network to use it whenever you want, there's no escaping much higher power draws, even at idle. Wherever electricity is expensive, the electric bill can be a prohibitive barrier.
butvacuum 15 hours ago [-]
No-Load losses on the power conversion seems to be the real issue and is rarely mentioned and understood even less often. Plus the conversion eff. curve.
Eg: I shaved ~40W off the idle load on a server (250->210W) by doung nothing more than removing the redundant supply
rmunn 13 hours ago [-]
You just reminded me of a funny story about redundant power supply.
A college buddy used to work at Motorola (I'm naming the company because they wouldn't mind this story being shared) back in the late 90's or early 2000's. They had redundant power to their campus, bought from two different companies, coming in on opposite sides of the campus, so that even if some backhoe operator cut a ground-based power line somewhere, they wouldn't lose power.
And yet, one morning, the power went off all across their campus. After a little investigation, they sent pretty much all their employees home at noon and told them "take the afternoon off, don't come back until tomorrow, you wouldn't be able to do any work anyway". Turns out that although the power lines came in at opposite sides of their campus, somewhere a few miles away both of the power lines feeding their campus ended up running through the same underground conduit. And yes, a backhoe had managed to cut that conduit and break both of the lines they depended on at the same time. They had a single, VERY non-obvious, point of failure, and the backhoe had unerringly homed in on that SPoF.
pixl97 15 hours ago [-]
Smaller solar setups might become a hedge against high power bills in the future it seems.
BeetleB 16 hours ago [-]
> Unfortunately, I can't not use AI because it provides the best speech recognition I've ever used and, going along with hand problems, it now allows me to write code after a three-year absence from that skill.
I'll second this. The combination of Whisper + LLM makes speech recognition fantastic. I occasionally have arm pain from typing, and this is a Godsend.
I don't use it to write code - but in my experience stuff like emails + docs was the greater source of pain (one generally types slower while coding).
globular-toast 6 hours ago [-]
You can not use it. You spent your life minus a couple of years not using it. You could totally get back into coding but LLMs have you trapped like a tethered elephant.
folkrav 40 minutes ago [-]
Missed the part where they had stopped entirely because of chronic pain? Are you somehow aware of the levels of pain they deal with daily to be able to say they could “totally” do it without?
globular-toast 18 minutes ago [-]
They didn't say anything about chronic pain, but with that extra context I do understand now.
1718627440 1 hours ago [-]
First, I am kinda the same.
> how do I do this one piece of a query in MySQL, since I'm not a MySQL DBA
Honestly for most things, the actual documentation is the best resource. I needed to learn to stop looking for solutions to my issue on various fora (like SO) and just read the primary documentation from the vendor. Have a look at dev.mysql.com or mariadb.com/docs/server
Schlagbohrer 1 days ago [-]
Amazon has implemented a password system for RAM delivery. You're supposed to give a password to the deliverer before taking delivery of the RAM, and you're not supposed to send the password over the phone at all (although the Amazon delivery guy still called me and asked for it over the phone).
Aurornis 17 hours ago [-]
Depending on your location this will happen for any item with a certain value and presumably risk score. It's not a unique decision they made for RAM. Probably happened to your order because the value was high enough and your location, not because it was RAM specifically
m463 17 hours ago [-]
I've already been burned via amazon a few times recently.
Most recent one I bought a multi-tb hard disk, but got an old multi-gb disk shoehorned into legitimate box. Couldn't just get a replacement - had to buy one for more money since that drive's price had increased.
Roast 1 days ago [-]
This is quite common for all high value deliveries across all platforms in India.
prodigycorp 1 days ago [-]
can u provide some context? what is this trying to solve and how does it solve it? very curious.
codeulike 1 days ago [-]
Its a way to make sure the delivery takes place. A code is sent to recipient and recipient only gives it to the delivery person once item is handed over. Delivery person cant mark item as 'done' without the code. I've seen Uber Eats do the same thing sometimes.
apple4ever 22 hours ago [-]
Wow that's actually a really good idea to prevent delivery theft by the delivery person!
stingraycharles 1 days ago [-]
What’s wrong with a signature?
supersparrow 1 days ago [-]
Signatures are completely pointless. Literally draw squiggles as anyone else can. Nobody ever checks or compares them to the original, how can they?
cromka 6 hours ago [-]
You could also claim it wasn't your signature and be refunded, but you can't claim that if the delivery person obtained the password only you had.
brewdad 9 hours ago [-]
My signature on a touchpad is noticeably different from my paper signature. It’s been a real challenge for me recently as I’ve had to sign a number of paper documents while the “official” signature on file for me is an electronic one from a touch interface. A few times I have had to resubmit fresh copies when the first ones were rejected.
jacobwg 1 days ago [-]
It's to prevent theft by the delivery person or by the wrong recipient
Ekaros 1 days ago [-]
Sometimes I wonder if true reason behind them is not any sort of verification but to maybe increase potential punishment if caught... So effectively nothing...
JuniperMesos 6 hours ago [-]
Increasing potential punishment if caught - or making it more likely that you will get caught - dissuades thefts.
pjc50 1 days ago [-]
Signatures are terrible, and especially it's hard to automatically confirm they're correct even if you have a sample on file.
vee-kay 1 days ago [-]
[dead]
beckford 17 hours ago [-]
They did the same for a Samsung Fold phone delivered to me last month (in the US).
17 hours ago [-]
preisschild 24 hours ago [-]
This is normal for every item above ~100EUR delivered by Amazon here in Austria. Never had it happen that they asked for the code without showing me the package in-person first though.
varispeed 16 hours ago [-]
This is a shame, because now I cannot order from Amazon as orders go to building reception and nobody is going to remember the pin to take delivery.
i feel like 90% of complaints about ram shortages i've read here could be solved by realizing that we mostly update devices for status reasons
we would collectively all be better off by being forced to stick to older devices, value software optimized for proper resource management, and learn that instead of reflexively accepting "problems" we can see them as choices
enjoy the rat race
j1elo 2 hours ago [-]
I'm ironically using AI to program an extra-lightweight 10-foot UI for an HTPC that's as underpower as one can be, and KDE Plasma BigScreen trashed it hard.
Reason being, a more powerful one would cost more than what I'm willing to pay (and, mostly, for fun and learning, which is priceless)
cmrdporcupine 1 hours ago [-]
Me, I bloody well love optimizing and performance tuning software. And love "retro" hardware. I'm fully on board with this world of squeezing every last drop of juice out of older systems. It's fun.
I have a minion in a terminal grinding away optimizing CPU-only LLM inference on Dual Xeon E5s. Why? Because it's neat.
tacocataco 4 hours ago [-]
Why stop at 2027? Why not just declare that you get everything produced by humanity for infinity years?
How is a new business going to compete with incumbents that buy 100% of the necessary components needed to enter the market? Is no one else allowed to compete in this market anymore going forward?
pizzaiolo 3 hours ago [-]
It's not permanent. It's only this way until the bubble pops.
inigyou 3 hours ago [-]
With infinite money printing will that ever happen?
embedding-shape 3 hours ago [-]
Infinite money printing always been possible, all previous bubbles eventually burst, that's why we now identify them as bubbles in the first place.
dsign 16 hours ago [-]
Hmm. Forget about desktops. I'm programming a microcontroller which is actually reasonably priced and has gotten all its 32Mb of RAM in the same die as the processor, but I'm getting nervous and having a strong urge to stockpile...
jacksons5f 15 hours ago [-]
Well, that one is likely out of danger - so far PSRAM (I assume that's what you have) is not impacted that much. However, we had to cancel some projects because we couldn't secure enough chips which have DDR on die at a reasonable price. They are 4x now compared to December last year.
ghtbircshotbe 1 days ago [-]
The boom cycle looks like it's going to last a while, but it never lasts forever. If the current memory manufacturers all conspire to not build more capacity, I'm sure someone else will (eg China).
inigyou 17 hours ago [-]
CXMT is now producing as much as one of the big three. Only for China, of course.
odo1242 15 hours ago [-]
They are producing about a third of what Micron (smallest of big three) produces, and plan on being on par with them in ~4 years.
dabinat 14 hours ago [-]
That’s less that China needs to buy from the rest of the world, so it’s still a win.
They've been caught as a cartel before, a leopard never changes its spots, so its going to happen again.
1970-01-01 15 hours ago [-]
Seems like we need a USB equivalent standard to RAM sticks. I have lots and lots of old (1-2GB), intercompatible sticks sitting. Sometimes you just need the space and not the speed.
haroldp 14 hours ago [-]
> Sometimes you just need the space and not the speed.
Your computer already supports that. It's called "swap".
OJFord 14 hours ago [-]
Yeah but DDR3 was faster then and still is now - I can't imagine my M4 Mac having a slightly older version RAM would meaningfully affect me, but another 24GB of it (or whatever I have) would certainly be an improvement.
CPUs have L1, L2, L3+ caches too after all.
icedchai 11 hours ago [-]
It probably would. LPDDR5 memory bandwidth on M-series Macs is several multiples of the memory bandwidth available on DDR3 (depending on the exact chip, Pro, Max, etc... 4x to beyond 10x.) The whole system would be much, much slower.
haroldp 11 hours ago [-]
> Yeah but DDR3 was faster then and still is now
USB would be the limiter for both.
OJFord 45 minutes ago [-]
I didn't say anything about USB? If you're referring to upthread wanting RAM to be standardised like USB (not standardised on something like USB), I took that to mean backwards compatible slots. (And maybe handicapped forwards compatibility, though why would you.) M.2/PCIe is this way.
__david__ 10 hours ago [-]
They obviously did not mean not literal USB, just a standardized connector across DRAM versions/speeds.
haroldp 8 hours ago [-]
Yeah, sure but... if it's real fast, lets connect SSDs to it and have real fast swap. Whatever it is will be limited to the bus speed. And that's gonna be silly compared to RAM speed.
astrobe_ 7 hours ago [-]
You don't need swap, you need "apps" that take back the "memory is cheap anyway" crappy culture.
Dylan16807 13 hours ago [-]
A better SSD can get up to a passable GB/s rate, but when you're waiting tens of microseconds instead of tens of nanoseconds you can't use it the same way. And a worse SSD is 10x slower in GB/s too, compared to old super cheap RAM.
undersuit 15 hours ago [-]
Grab the cheapest Intel Optane M10 drive from eBay with a 1x PCI-E adapter from them too. 16GB of Swap for less than $20.
The M10 are over produced and only PCI-E 3.0 2x lane devices but 16GB of DDR5 is going to cost over $200.
You can do the same with USB-C M.2 adapters but the adapters cost more and eww USB for Swap, hope the device never drops.
hdb2 14 hours ago [-]
Hey, I apologize for using you as tech support, but I found tons of those M10 sticks you are talking about, but what should I be looking for in terms of an adapter? All I can find are adapters listed as M.2:
I believe the intention is for you to put it in the M.2
Melatonic 11 hours ago [-]
Is that really competitive with just doing a pciexpress card with faster M2 SSD ?
undersuit 10 hours ago [-]
Optane on a cheap adapter is superior in just this Swap application because the 16GB M.2 are very cheap, the 2x lanes aren't under utilizing your 4x motherboard M.2 slot when used with an adapter, and Optane has better write endurance and latency then even the newest flash.
Melatonic 9 hours ago [-]
Read more about it. Not sure on endurance but does it really do that well on latency ? Curious if you have any resources. Some boards have a lot of spare m2
I suggest using an 1x PCI-e adapter because putting a M10 in a M.2 slot is likely going to only have 2 of 4 PCI-e lanes in use. You have to figure out the tradeoff of using a Swap on your, now very expensive as well, Flash storage that may have flashier sequential read and write speeds but doesn't have the ability to sustain random read and write performance which is what a Swap is going to use the most.
There are faster Optane drives that use 4 PCI-e lanes or even PCI-e 4.0, but they get pricey quickly, so you should just buy ram then. The deal is the 16GB Optane on an unused 1x PCI-e slot.
Melatonic 8 hours ago [-]
What about using one of those fancier enterprise Optane drives (that are also larger) either as a swap or - dedicated drive for writing or reading huge amounts of small files ? The Intel P5800x is the opposite of cheap but from tests I'm seeing the latency and performance is very impressive for some niche use cases
jurf 15 hours ago [-]
RAM over “USB” seems dangerously close to an SSD.
In any case DDR3 isn’t nearly as bad price-wise and would have much better energy efficiency than stacking 1–2 gig sticks.
ramraj07 13 hours ago [-]
Thats like saying I need a car with hotswappable engines. Sure. But you wouldn't want that car anymore.
For starters, The slowest sticks of DDR(x) are often slower than DDR(x-1). The issue is never capacity, but rather, performance.
The real nonsense? USB? It's a mess. Pick a USB cable and buy it from ANY retailer, let's make it easier, buy a USB-C cable. What is the data rate (depends on cable quality and length), Does it support power delivery? If so, how many watts? (~5W requires a very different cable from ~230W), how do you know from simply looking at the cable? If you buy a cable, how can you tell what it supports by simply looking at the connector? Imagine having a box full of USB-C cables. Could you tell me how fast each of those cables are? (The cables themselves don't! Many don't have any markngs at all, and if they do, the markings could be fraudulent)
RAM does not have that issue. A stick fits or it doesn't. Sure, there are a huge range of speeds, however, that range has a ballpark (JEDEC defines the ballpark, the "cartels" make the memory and push out some faster stuff).
While there are definite exceptions (I was bitten by one recently), you can generally plug in a DDR5 DIMM and expect it to work in the system.
The same cannot be said for USB. Some USB cables ONLY deliver power. Some only work with certain devices. I've USB-C (!!!) cables that only work with the devices they are shipped with, and even more annoyingly, the both may be true! ASUS (!!!) ships MOTHERBOARDS that can only pair with vital hardware via very specific USB-C cables (Strix Hive) and "GLORIOUS" ain't so glorious. Their mouses warn you not to use other USB-C cables, and they are right, depending on the make/model/generation, you can brick your "glorious" peripheral.
No, USB nonsense needs to stay far away from anything else.
Also, the reason this is a huge issue is because the memory makers are cartels. Only a few of them exist, they gang up and bully EVERYONE and nobody has ever invested money to create a competitor due to this, except China, which of course means that the U.S. and portions of the E.U. are insta-banning/trying to insta-ban, even though it is really freaking hard to install spyware on a memory module.
seba_dos1 14 hours ago [-]
> Pick a USB cable and buy it from ANY retailer, let's make it easier, buy a USB-C cable.
Okay, let's assume it's a USB-C cable and not just something USB-C shaped that pretends to be one.
> Does it support power delivery?
Yes.
> If so, how many watts?
Always at least 60W (3A). Up to 240W if it's e-marked.
> What is the data rate
Depends whether it's a USB 2.0-only cable (480Mbps) or a complete cable (20Gbps). Possibly more in Thunderbolt or USB4 modes - cables that handle those will be marked (both visually and with an e-mark). In any case, pretty easy to check just by plugging things in.
YuukiRey 5 hours ago [-]
But how do I know if a random cable that comes with a device I bought is just something USB-C shaped or the real thing? How do I know the same thing on Amazon or whatever other big retailer with thousands of cables I use?
The last time I bought one I looked for one with a proper certificate but that eliminated the vast majority of cables on Amazon and I suspect that I ended up buying something ridiculously over engineered.
numpad0 3 hours ago [-]
No one explains it this way for some reason, but USB-C "60W" is more like USB-C 20V. All USB C-to-C cables are supposed to be rated to something like 600V by laws and common sense, and 3A by USB-C spec. 20V x 3A = 60W.
USB-C in C-to-C configurations require all the PD pins working for voltages to be supplied at all. So all USB-C cables that aren't the ones that actually don't work at all should be able to do that 20V x 3A -> 60W. Only in rare cases where a C-C cable is wired to make it look like one by being internally wired as C-A-A-C it becomes problematic since only USB-C has the config pins used for non-5V modes, but that should be actually rare.
I think USB-C shaped thing mentioned in GP refers to USB-C devices, like dollar store Bluetooth speakers and such. For those things, it's in fact not so rare that they come with tons of stupid quirks.
The can of worms is mainly in USB 3.0, Alt Mode, 5A mode, and Thunderbolt/USB4. 3A mode is fine. Raspberry Pi notoriously used a "5V 5A" mode that it's not supposed to be using.
Dylan16807 13 hours ago [-]
I don't know why you got downvoted so much. There's no such thing as a 5 watt USB cable from a reputable source and people need to stop acting like there is. 60 is a very good baseline. There's no such thing as a power only USB cable either.
If we're counting fake products then we need to count those RGB sticks that have no RAM in them.
seba_dos1 11 hours ago [-]
You couldn't even make a 5W USB-C cable if you tried. You'd have to make the wires so thin that it would burn at above 1A I guess, and it would have no way to signal the device to limit its power consumption to 1A, so it would just burn once you tried to use it. All you need to handle 60W is to make three electrical connections between the ends, and the cable won't work at all if you miss any.
TZubiri 14 hours ago [-]
USBism has a lot of elements similar to the "connect your agent to your data with this 'curl | sh' script and do anything. Maybe some friction and incompatibility is useful in some cases.
tosti 1 days ago [-]
All those devs who were saying "ram is cheap" BTFO.
Pretty sure those devs were saying the same thing way before 2017 as well, which seems to be ~ the last time RAM was this expensive.
Now RAM in the cloud, now you're really paying the Java premium.
tosti 22 hours ago [-]
Price per GiB is skewed by definition. 10 Years ago, 8GiB was enough and cost between $10 and $20. For the longest time this century, a cheap build was in the $250 to $350 range. Nowadays I'm having trouble with that price range building from used parts.
adrian_b 1 days ago [-]
According to your link, DDR5 today has the same price in real US $ as DDR3 was in 2014, during a brief spike towards high prices.
Since that peak in 2014 until today, the prices were lower than now. The peak in 2017 was more than 10% lower than today.
To reach permanently higher prices than today, we must go backwards until 2011.
So we have already regressed at least 12 years into the past, but more likely 15 years, and it is unknown how much more we will regress.
zahlman 15 hours ago [-]
> we have already regressed at least...
It seems worth noting here that DDR5 RAM is faster than DDR3 RAM, and was not even available until 2020.
jbverschoor 1 days ago [-]
Thanks to AI, we can now one-shot convert electron to native apps ;-)
gaigalas 1 days ago [-]
Recipe for delicious pizza:
1. Make extra greasy pizza. Nauseating-level of grease.
2. Put pizza through De-Greasinator 5000.
Sounds efficient.
inigyou 1 days ago [-]
The grease is vaporised and inhaled by one randomly chosen Memphis resident, who permanently gets chronic asthma.
Dylan16807 12 hours ago [-]
You only need to make the app once before running it on a million computers. So it's more like greasy flavor development, then de-greasinating the recipe before going into mass production. It's not a big efficiency impact overall, it's more about how you like to spend your flavor development time.
WD-42 13 hours ago [-]
Maybe they should start with Claude.
m463 17 hours ago [-]
"it converted my gambling app into a three-armed bandit!"
forgetfreeman 15 hours ago [-]
Yeah would someone like to explain why there's an octopus in the break room?
shepherdjerred 21 hours ago [-]
OK I’ll take the bait.
Were devs supposed to optimize RAM for a shortage that might come? I’m guessing you had enough foresight to stockpile RAM?
compiler-devel 18 hours ago [-]
What’s implied here is that devs were supposed to optimize for user resources, not for developer convenience.
IncreasePosts 14 hours ago [-]
Maybe we can have AI rewrite those programs and optimize for low ram usage
Cthulhu_ 3 hours ago [-]
Maybe; if it's a Js / electron app there's a lower boundary though. The main issue is that the web (html/css) is the most versatile UX platform out there, nothing even comes close. You can't build most web apps' unique visual style in any other platform. Yet. Let alone have it crossplatform.
theevilsharpie 38 minutes ago [-]
> You can't build most web apps' unique visual style in any other platform.
Every native UI framework I've ever used allows drawing directly to a canvas, and can have literally any visual style that can be created with pixels.
pessimizer 17 hours ago [-]
They were supposed to optimize for the situation in which they weren't the only program running on a computer, and they failed to do it.
Pavilion2095 17 hours ago [-]
Can you blame them? Users didn't care. Heck, developers didn't care - look at how popular vscode is.
glaslong 17 hours ago [-]
Revealed preferences aside, every dev comment section for 10 years HAS been full of people loudly insulting Electron.
odo1242 14 hours ago [-]
Honestly, despite being Electron VSCode manages to outspeed every other editor on my machine (by qualitative metrics), even stuff like Zed (which manages to lag on stuff VSCode doesn’t despite performance being its main selling point)
SoleilAbsolu 16 hours ago [-]
Time to bring back 32 bit OS, that way you only need 4GB RAM per computer!
vlod 10 hours ago [-]
640k ought to be enough for anybody. Sorry had to be said.
amelius 1 days ago [-]
Bad times for the two-space copying garbage collectors.
pjc50 1 days ago [-]
Are there any of those left? The popular VMs for Java, Javascript and dotnet are all very mature heavily tuned designs.
GoblinSlayer 23 hours ago [-]
My benchmark:
1) .net app, one text field, one button; private bytes 22mb, working set 27mb
2) native app, two text fields, three buttons; private bytes 1.2mb, working set 7.3mb
tosti 22 hours ago [-]
3) electron, about:blank, 1.2GiB
adzm 16 hours ago [-]
electron, about:blank, 90mb
Just tested
hombre_fatal 9 hours ago [-]
That's better than the GTK4 hello world I tried out last year.
odo1242 15 hours ago [-]
Seconded, the only 1.2 GB electron app I’ve made had a ~500MB ML model’s weights embedded in it
nixon_why69 1 days ago [-]
They're heavily tuned but I think having the generational hypothesis and copying longer-lived objects is still standard? As is allowing a fair amount of garbage to accumulate to increase throughput and reduce copying at the cost of memory footprint?
Posting instead of researching in hopes someone smarter can chime in, because I'm lazy.
rock_artist 1 days ago [-]
That's tech... Moore Law was steadily going until raw CPU got stuck for a few years, switching from single core to multicore then heterogeneous cpus, all those requiring to rewrite software in a scalable manner and then consider newer schedulers switching between performance/efficiency cores and changing CPU freqs.
Also, you have Meta and Google investing in simplified performant versions of their stack for developing countries which is similar.
What's funny though,
I see people that sometimes says it's cheaper to build the app they need in a single prompt than to search for it.
eru 1 days ago [-]
You know that Moore's law is about transistor count? What worked really well for CPU in eg the 1990s was Dennard scaling. That drove the Hz higher. And Dennard scaling is what stopped working.
Moore's law is alive and well.
dgellow 1 days ago [-]
It also was just an observation, not a strict rule set in stone
otabdeveloper4 11 hours ago [-]
Moore's "Law" is not a law, it's just an observation of market forces at a particular time.
chvid 1 days ago [-]
I saw on Twitter that TSMC has made 1B worth of chips for Apple that they cannot package because of missing memory.
(PS: It is time to let ASML sell to CXMT.)
HarHarVeryFunny 16 hours ago [-]
I thought CXMT already have plenty of capacity - Apple are currently talking to the US government for blessing to buy from them.
Other manufactures like Micron do have aggressive expansion plans, but fabs take a long time to build and bring online - multiple years.
Don't really need the EUVs for commodity DRAMs and China has plenty of DUVs capable of 10nm. CXMT's DDR5 is known to use 1z nodes (17nm) vs 1c (10nm) used by the big three -- ie, higher cost, less efficiency, less density.
The big three are also expanding their capacity. IMO, it seems to make more sense to sell equipments to those who could make better, efficient use of them. ASML has a backlog of €38.8 billion and is fully booked for 2027.
vdfs 1 days ago [-]
> (PS: It is time to let ASML sell to CXMT.)
Why? High demand and low supply is very good for shareholders and make indexes go up, their goal is not to meet demand, they want exponential qurartly revunes
inigyou 17 hours ago [-]
You still have to actually meet the demand. It's not good if you don't sell any products.
stingraycharles 1 days ago [-]
It’s temporary revenue, it’s not as if those businesses are suddenly fundamentally better.
To me it seems like they’re just riding the AI hype out, expecting a crash after, and not willing to risk a lot of capital on someone else’s bubble.
Foobar8568 17 hours ago [-]
Temporary revenues for decades to come.
stingraycharles 14 hours ago [-]
That’s very much up for debate.
chvid 1 days ago [-]
True - but what is happening now is very destructive to a lot of important companies.
ethbr1 1 days ago [-]
Indeed. There's a reason the US government has historically stopped air traffic / railroad / longshore strikes: cascading effects.
If memory is now upstream of so many companies, then maybe a higher regulatory involvement would be appropriate (like power companies).
Specifically, and it could fix all of this, capping the max % of their supply they're allowed to sell into the AI space.
nixon_why69 1 days ago [-]
I hear what you're saying but memory ultimately is not infrastructure, it's a commodity that's shipped over infrastructure.
The arguments for infrastructure being managed more tightly by the government is because it's a natural monopoly that everything depends on. Memory isn't a natural monopoly, if we found ourselves in a spot where we all depend on 3 companies then that's not a law of nature, we need more companies.
0cf8612b2e1e 16 hours ago [-]
The government interferes with many commodities. Food, oil, rare-earth elements, sugar, airplane frames, etc
preisschild 24 hours ago [-]
Its not like demand will be lowered by this though, I dont see how this would be bad for ASML (besides PR-China generally not respecting IP and them reverse engineering it)
znpy 24 hours ago [-]
I just thought… didn’t Intel start as a memory chip company?
Couldn’t intel repurpose some of its fabs (maybe the old ones?) to produce memory?
I’m no expert but… seems like an easy win?
inigyou 17 hours ago [-]
You want a bunch of 4 kilobyte DRAM chips with an interface that predates DDR1, for your Commodore 64?
rgmerk 2 hours ago [-]
I dunno, a RAM expansion on my Commodore 64 Ultimate might let it open bigger documents in GEOS…
MotoriX 2 hours ago [-]
I’m so glad I managed to upgrade my PC at the end of 2025, literally just a week before DDR5 prices started going through the roof...
amelius 1 days ago [-]
How am I going to buy a computer that connects to AI datacenters now?
glaslong 17 hours ago [-]
I believe you're just supposed to stream your video and audio up through your Meta Glasses.
Or lease a laptop from Apple if faceputer's aren't as much your style.
cobertos 1 days ago [-]
I'm sure some enterprising company will find a way sell you a computer subsidized by your data usage or maybe by forfeiting your first born
tech2 3 hours ago [-]
All that's old is new again. In the late 90s there were a number of companies (eMachines, for example) which heavily discounted their machines by having integral advertising, and in some cases through internet service provision. This sounds like the same idea all over again, just with a different backing fund source.
HarHarVeryFunny 16 hours ago [-]
Time to go dig an Apple II out of the landfill
RiverCrochet 15 hours ago [-]
From the "Apple II (original)" Wikipedia article:
"The original retail price of the computer with 4 KiB of RAM was US$1,298 (equivalent to $6,900 in 2025)[21] and with the maximum 48 KiB of RAM, it was US$2,638 (equivalent to $14,020 in 2025)"
ZoomZoomZoom 1 days ago [-]
Don't worry. Sooner or later this stuff will be mandatory and you'll be obligated to get your state-issued terminal. Social credits will do, if you have no money.
For the skeptical ones: Somewhere people are already being cut off from state and financial institutions without proprietary software with bundled security certificates, applications are bound to collect data on the environment they are used in (other apps) and organizations deliberately limit access to their services or cripple them without their applications (can't do things from a generic web-browser).
zahlman 15 hours ago [-]
> For the skeptical ones
The words after this point make you less convincing, not more, as they only serve to confirm the egregious slippery slope of your reasoning.
ZoomZoomZoom 13 hours ago [-]
Do you call any negative prognosis based on extrapolation of an observable downward trend a slippery slope?
techteach00 16 hours ago [-]
Yes and if it's true it's usually downvoted or flagged on HN
pickleglitch 23 hours ago [-]
Jeff Bezos will be happy to let you rent one from him. He's said this is how he envisions the future of personal computing. You will own nothing, and so on.
ppierre 3 hours ago [-]
Wait of few months: "AI companies can't pay"
krm01 3 hours ago [-]
This happened once before (less dramatic but still) until at some point the prices came down crashing. I can’t see this holding for much longer.
tripleee 5 hours ago [-]
next gen gaming consoles are going to be at or over $1k. Honestly I think they're dead in the water in these conditions
pjmlp 2 hours ago [-]
Great, maybe we get back proper native applications.
GMoromisato 14 hours ago [-]
I just bought 16 GB DDR4 for my (somewhat old) Dell tower for $120 + shipping from Techyparts.com. Looking around that seemed like a really decent price (but maybe I'm wrong).
I had previously ordered from Aeonfly, but they cancelled my order--likely because they wanted to raise the price.
wbsun 6 hours ago [-]
I guess very soon we will see on HN that lots of young folks will be excited )share their findings of some “brilliant” and “innovative” ways to save memory.
rgmerk 1 hours ago [-]
I’ve no doubt they’re setting up memory-footprint auditing agents as we speak.
Szero0x00 27 minutes ago [-]
.
WithinReason 1 days ago [-]
This hopefully helps mitigate the risk of putting more production capacity in place, which the Big 3 were not willing to do so far.
pjc50 1 days ago [-]
At this size, I think we have to start worrying about a 2008-type default risk. AI companies will have ordered all the RAM and paid some kind of deposit for that, but actually taking delivery and paying for it is still in the future and the sums of capital involved are extremely large.
nixon_why69 1 days ago [-]
Also there is a ton of debt flying around, its not just equities hurting VC funds, real banks are facing risks now.
inigyou 1 days ago [-]
That's the reason RAM companies aren't expanding. Worst case, they can sell their normal production to gamers.
londons_explore 1 days ago [-]
Fail to deliver, and they get bankrupted...
Do deliver and they don't get paid because the buyer purchased with money that doesn't exist.
Lose-lose situation!
inigyou 1 days ago [-]
But they don't get bankrupted, because it's an ordinary year for them, only with different customers.
londons_explore 1 days ago [-]
Fail to deliver, and they get bankrupted because a court would not simply unwind the deal - a court would demand compensation of the market price of the goods not delivered on delivery day - which will probably be very high.
1 days ago [-]
eru 1 days ago [-]
What kind of default risk?
What we got in 2008 was central banks (eg Fed and ECB) willfully collapsing nominal GDP in their economies. Have a look at the dot-com bust or Black Monday for comparison.
pjc50 1 days ago [-]
> willfully collapsing nominal GDP
What on earth are you talking about?
The credit risk was a very real problem; Kaupthing, Anglo Irish, RBS, Lehman etc.
eru 1 days ago [-]
We just had a string of bank failures, like Silicon Valley Bank, but you will not see any impact in the macro-economic data (like unemployment), because the Fed kept nominal spending on track.
In 2008 they didn't. Instead they actively tightened monetary policy by eg introducing interest on excess reserves. The ECB even hiked interest rates.
Companies defaulting on debt doesn't need to bring down the economy.
For the US, you can also see how the construction sector had been winding down for years (eg as measured in construction employment) without an impact on overall unemployment. The crisis was entirely avoidable.
AI companies would do well to realize that civil disobedience is a thing.
Take what's currently happening to Flock cameras as an object lesson. People are so fed up with Flock that they're cutting their cameras down en masse, sometimes wiping them clean from entire jurisdictions. Tens of thousands of other people in those jurisdictions are cheering them on, even offering alibis for them before they've been caught. Law enforcement doesn't seem too keen to catch them either. You absolutely can be so hated that the law will not be enforced to protect your property.
If AI companies don't slow down and build some good will, AI data centres are probably at risk. They'd probably have been targeted before Flock were they not harder targets. A higher degree of cooperation and organization will be required to sabotage them, but it would be foolish to believe it won't eventually happen.
Monopolizing memory production for another year suggests that lessons have not been learned, the mad build-out will continue, and we're headed for some truly crazy stuff.
apple4ever 22 hours ago [-]
That's a really great point. So much about our world is about perception. I think the datacenter concern is overblown for the most part (except putting them next to houses), but the perception is stronger than the facts.
And if AI is putting people out of jobs and making everything expensive - or that's just the perception - that's gonna set AI back for years.
fwlr 5 hours ago [-]
I’m pretty sure this is why the “data centers in space” idea just won’t die. Yes, it has so many disadvantages, but it does have one distinct advantage of being outside the Molotov-throwing range of the public.
timedude 12 minutes ago [-]
I think those space datacenters still have control centers on earth
swed420 1 days ago [-]
> Take what's currently happening to Flock cameras as an object lesson. People are so fed up with Flock that they're cutting their cameras down en masse, sometimes wiping them clean from entire jurisdictions.
It seems inevitable that these lost cameras will simply be upgraded to drone fleets. Not an identical replacement, but still "good enough" for the original purposes. Had they anticipated this problem and skipped the first phase, it might have been harder to manufacture consent for privacy invasion. But now anybody who was okay with the original cameras feels like a victim having them torn down, which makes it likelier they'll be okay with drones everywhere.
> Tens of thousands of other people in those jurisdictions are cheering them on, even offering alibis for them before they've been caught. Law enforcement doesn't seem too keen to catch them either. You absolutely can be so hated that the law will not be enforced to protect your property.
At least one third of the country is still okay with them, because they believe it helps rid them of illegals and deter crime, with downsides small enough to ignore, in their mind.
In sum, we don't have as much control over these things as you seem to believe. Capital was always in control with its crude but effective propaganda pipelines to manufacture enough consent. Now imagine those pipelines becoming even more effective thanks to AI. Not looking good.
customguy 21 hours ago [-]
> anybody who was okay with the original cameras
Who cares about those people? I would argue even they themselves hardly do. And sure, they'll be even more okay with even worse stuff, and that even worse stuff will be even more hated by even more people, whatever the people who are "okay with" (that can mean ignorance, obedience, or evil intent, but it's not justification). If they can be okay with mass surveillance they can be okay with other things, so they're basically a wash, they're just okay with stuff.
> At least one third of the country is still okay with them
Then they can be okay with the majority saying you know what, we're sick of you, you can either stop having your toys or you will be segregated from us and live under your own surveillance. "Being okay with something" isn't some magical wand or a fortress, it's just a limp shrug. It means "pass".
> In sum, we don't have as much control over these things as you seem to believe.
I don't see the connection with the sentence before that, which is a minority being "okay with" something and a majority willing to fight and out there doing it. It's not about controlling what a minority of people are okay with, it's about changing what is done. The people who are okay with stuff can go read a comic book until it's over, how does that not solve anything you raised about them? Or are you saying their being okay with stuff should be respected or even heeded? Couldn't they just make that claim themselves, if they wanted to?
Imagine a raging house fire, people stumbling over each other trying to help, and some guy steps up and goes "WAIT!" they all look at him, sure he has a major contribution for interrupting something like this, a good idea, a plan perhaps; and he goes "... I'm actually okay with this."
swed420 20 hours ago [-]
> Then they can be okay with the majority saying you know what, we're sick of you, you can either stop having your toys or you will be segregated from us and live under your own surveillance. "Being okay with something" isn't some magical wand or a fortress, it's just a limp shrug. It means "pass".
It's not the majority, though. Roughly speaking, among adults, we have:
* 1/3 who are willing to vote for the red flavor of lesser-two-evils (some of them don't even register it as evil)
* 1/3 who are willing to vote for the blue flavor of lesser-two-evils (some of them don't even register it as evil, including when it literally manufactures consent for the first third to win next time around)
* 1/3 who see the above as a uniparty of capital interests offering the illusion of democracy through "close" elections as the bullshit that it is.
nancyminusone 15 hours ago [-]
>upgraded to drone fleets
I wouldn't do that unless you want your airspace filled with birdshot
nullbio 7 hours ago [-]
They don't need to slow down development, they just need to slow down on being greedy over all of the hardware. They should be investing in manufacturing.
8 hours ago [-]
lofties 11 hours ago [-]
> AI companies would do well to realize that civil disobedience is a thing.
Civil disobedience? In this economy?
kristopolous 10 hours ago [-]
If someone bombed a data center there would be celebratory parades.
Nobody wants this robber baron shit
14 hours ago [-]
amelius 1 days ago [-]
Except nobody knows where the AI data centers are actually located.
apple4ever 22 hours ago [-]
It doesn't matter. The perception is all datacenters do AI (not necessarily wrong nowadays either) and that's all that matters to the public when they see so many lost jobs and high computer / phone prices.
amelius 22 hours ago [-]
And how does the public know there is a datacenter inside a random building?
apple4ever 20 hours ago [-]
Either it's being proposed or other people tell them about them. It's not hard to find out where datacenters are.
skinfaxi 17 hours ago [-]
The local government would need to know, and that would come out in town meetings.
kps 16 hours ago [-]
Browse Google Maps aerial view and look for the backup generators.
lostmsu 15 hours ago [-]
Every produce storage facility going to have them, won't it?
prepend 13 hours ago [-]
No. It’s not economical for produce to have backup generators. Better to just move produce or let it go.
HWR_14 16 hours ago [-]
"That's not a datacenter. It's just a regular old building consuming 20+MW of electricity"
preisschild 24 hours ago [-]
You can hear them from quite far away since companies like X use on-site Methane Gas Turbine Generators because they don't want to wait for the grid capacity to catch up.
Buying and selling at market price is rarely illegal. I think in the USSR it might be?
HWR_14 16 hours ago [-]
In the US, selling at the (new higher) market price during a declared state-of-emergency is illegal in most states and territories.
NewsaHackO 14 hours ago [-]
Are talking about selling marked up water during a state of emergency? Don’t think that would apply here.
HWR_14 13 hours ago [-]
Water is one example, but it usually goes beyond that. I'm not sure if RAM is ever covered by those laws, but I'm also not willing to look through every state's laws and check.
chupasaurus 1 days ago [-]
It was legal in USSR for USSR to scalp though, depends on point of view.
keybored 1 days ago [-]
Ha ha, the USSR was a silly thing. Anyway, what was the topic about?
inigyou 1 days ago [-]
I think it was about the failures of centrally planned 5-year plans of RAM sales.
actionfromafar 1 days ago [-]
Isn't scalping a term used for tickets? Also, these companies supposedly intend to actually use the RAM they ordered, not reselling at some inflated price.
It's all fair game IMHO, except the elephant in the room which is that the AI company bubble might pop.
GoblinSlayer 1 days ago [-]
Then we'll have a lot of cheap RAM.
tencentshill 24 hours ago [-]
Not sure if HBM can be used as regular DDR.
dspillett 1 days ago [-]
> Isn't scalping a term used for tickets?
It is also a term use in general trading anywhere traders open and close their positions quickly to make profit off an upswing which might be in part caused by traders doing this. If done with inside knowledge it is illegal, in this case more commonly referred to as “front running” (which is what many suspect happened with numerous ahemfortuitous/serendipitous position changes around announcements of changes in the state of the US vs Iran situation).
> Also, these companies supposedly intend to actually use the RAM they ordered, not reselling at some inflated price.
As well as the companies ordering for the purpose of DC roll-outs and upgrades, there will definitely be some buying purely to sell at a higher price a short time later.
> It's all fair game IMHO, except the elephant in the room which is that the AI company bubble might pop.
[and with reference to the sibling reply to this: “Then we'll have a lot of cheap RAM.”]
I don't think it will pop with a sudden bang, and even if it is it won't affect these memory sales or probably those for 2028. The first real effect will be those using the AI tools suddenly finding themselves needing to pay a lot more for them as companies offering those services deal with reduced access to “new” credit from the circling investment pool shore up their fanciful accounting that way instead. The DC builds/upgrades that are currently planned will happen but more funded by end users and less by the investment cycle. New plans might not be actioned at the same scale, but it will take a while for that to affect component prices as there is a lot of lead time involved.
AussieWog93 1 days ago [-]
I don't see the AI bubble popping any time soon (as in, not in terms of physical datacenter buildout - stock prices could fall). Even if progress hits a brick wall, we still have multiple years of just expanding out the models we have to more people in more industries.
user43928 1 days ago [-]
And if anything, I believe progress has been accelerating.
A year ago the SOTA was GPT-5/Opus 4.1/Gemini 2.5 Pro, two years ago Sonnet 3.5.
Looking at Opus 5 for SOTA performance and GPT 5.6 Luna for a viable cheap alternative, AI is much more capable now.
Honorable mention: GPT 5.6 Sol on Cerebras, capacity limited to a few customers, is supposedly serving 750 tok/s.
Compared to 90 tok/s for non fast mode 5.6 Sol, 56 tok/s for Opus 5, and 190 tok/s for 5.6 Luna.
I am very curious about the next generation of models, GPT-6/Astra is rumored to launch still in August. Not sure what is the state of Anthropic's next Fable checkpoint.
If these models also deliver significant improvements, I really do not see how one could seriously still argue among the lines of AI being a scam, and the demand not being there to support the size of the investments.
kergonath 1 days ago [-]
> I really do not see how one could seriously still argue among the lines of AI being a scam, and the demand not being there to support the size of the investments.
I don’t think most people are saying it is purely a scam (well, some do but I don’t think they are to be taken seriously). What all these talks about circular financing and VC money are saying is that demand cannot sustain the sector long-term, not that there is no demand. People are certainly happy to pay say $20/month for whatever AI chatbot, but would they still pay if it were $200/month, which is closer to the actual costs.
There are several other details that point towards a unsustainable projections:
- measurable benefits from AI-ifying companies are nowhere near what is commonly believed. AI providers are hoping that they can keep the show going until the models are good enough, essentially faking it until they’ve made it, but that is not a given. It is also unstable because it is susceptible to change in public opinion.
- permits for new datacenters are not going to become easier to get as public opinion keeps turning against them. They will have to concentrate in friendly regions, which will add cost (more demand for the same location, plus interconnection for network and electricity, both of which can easily become bottlenecks).
- the electric grids are not ready for all those planned datacenters, so something will have to give. Building more and more on-site diesel generator in times where oil supply is so constrained and random is not very sustainable either.
- eventually the loans will come due and if earnings do not match there will be a, possibly severe, correction.
If we look at the historical example of the dot-com bubble, the crash was not caused by no demand. It was just caused by over-estimated demand and too much money going to a single sector of the economy. We still use the Internet, and it is still hugely important, but the correction was still severe and real people lost real money.
user43928 1 days ago [-]
Those are good points, but some arguments are weak:
- Chatbot usage in $20 plans is likely nowhere close to a $200 cost.
Inference cost has come down rapidly, with reports from July claiming OpenAI can now serve all of the logged out ChatGPT traffic on just a few hundred GPUs.
- Measurable benefit: I doubt anything of value is being measured. MS Copilot with GPT 5.5 Instant processing SharePoint files? Developers using AI as a fancy autocomplete under a "I review every line" regime?
I rely on my personal value judgement, based on 30h/week I spend using AI outside of my regular job.
I am developing a mobile app, competing with companies with millions in revenue and entire dev teams. I know it is viable. Others lag in effective adoption, their opinion is likely to change soon with even more capable models.
- data center projects in the US: They look to me to mostly be constructed in the most remote backwater. If even there projects with such moderate environmental impact cannot be realized, that would be an embarrassing policy failure
- the grid: I think that one is true. AFAIK the constraint would be gas turbines, not diesel, and oil supply is not structurally constrained
- the loans: Anthropic is rumored to have become profitable earlier this year because of large growth in enterprise revenue. It does not look so bad to me
AussieWog93 1 days ago [-]
>I am developing a mobile app, competing with companies with millions in revenue and entire dev teams. I know it is viable. Others lag in effective adoption, their opinion is likely to change soon with even more capable models.
To add to that, as an ex-software dev that mostly works in a semi-unrelated field now (and can only code as a small part of my job): I think a lot of software devs are underestimating what someone with reasonable technical skills and specialised domain knowledge is able to create with AI.
I honestly wouldn't be surprised if, in 5 years' time, the majority of software used by (e.g.) potato farmers was primarily created by other potato farmers. The code might still be less elegant but I think it will be easier for the potato farmer to iterate with an AI than outsourcing to a dev firm.
user43928 1 days ago [-]
On a tangent, because you mentioned five years for widely used software being implemented by domain experts rather than software engineers: this seems like a very conservative timeline.
I have little doubt that a competent potato farmer could possibly implement and sell such software today using Opus 5, Fable, or 5.6 Sol.
Now what I am going to say next may sound a bit crazy, and it is outside of my area of expertise. I hear the recent mathematical breakthroughs made by GPT-6/Astra are field medal worthy discoveries.
What if the current trajectory of improvement holds for another year or two?
Is it impossible that LLMs gain superhuman ability to reason over a large number of constraints so that they can make novel breakthroughs unimaginable to us today?
What I hope to see in five years is not potato farmers writing software, but programmable immune cells that safely kill cancer.
Perhaps I am an optimist.
nullbio 7 hours ago [-]
We need more manufacturing capacity, or this will never end.
cafebabbe 6 hours ago [-]
I can think of a few scenarios where this will end brutally
1 days ago [-]
14 hours ago [-]
axegon_ 16 hours ago [-]
Yup... And in 2021(5 years ago now) I got 64gb ddr4 for 120 bucks, second hand with a 6 month warranty. I felt like I hit the jackpot then, now that price seems like literal peanuts. I can't wait for the AI bubble to burst.
narrator 15 hours ago [-]
All of you that are excited for the m5 mac studios are going to be so bummed.
dartharva 9 hours ago [-]
Let this be a wake-up call for self-touted "sovereign" economies around the world. Domestic fabs are a must.
1 days ago [-]
thomasjeff1 13 hours ago [-]
We are hammering our foot by investing in AI at this point.
AI grows and grows and becomes god and kills us all.
How did it all start? Oh we invested all our money in AI.
jgalt212 13 hours ago [-]
Queue the blog posts:
We rewrote all our Python code in Go because the RAMs too damn high!
brador 1 days ago [-]
China RAM entered the US market in quantity a week ago west coast. It’s slow buggy trash but it’ll cool the market until failure rates are known. Europe got its first shipment 3 weeks ago.
livinglist 6 hours ago [-]
Where did you get this info? What are these RAM used for? Data centers or consumer electronics?
0xbadcafebee 15 hours ago [-]
It was already reported that new RAM production wouldn't be available until at least the end of 2027, and when it does come back online there'll be a backlog of orders, so prices and availability won't be stable until at least 2029. That's assuming there aren't more global economic and political issues delaying it.
vonneumannstan 16 hours ago [-]
But Micron is overpriced and new capacity is about to come online right? Right???
globular-toast 6 hours ago [-]
I just have to hope none of my equipment fails in the next few years, I guess. I don't need or want to upgrade, but some of it is getting a bit old. I have spinning rust at the 6 year mark with no failures yet...
Part of me also thinks we need to stop this. How can we have a democratic society when only billionaires have access to computing equipment?
znpy 24 hours ago [-]
Honest question: is this the first time this thing happens?
Besides the AI craze, to me it sounds kinda reasonable. Most large buyers (apple, public clouds, OEMs) do forecasting and orders in advance, right?
trynumber9 17 hours ago [-]
Memory prices are cyclical. But this is bigger and longer than most cycles.
inigyou 17 hours ago [-]
According to semiconductor industry stock market analysts, memory gets really expensive about once a decade on average so far.
tonyhart7 8 hours ago [-]
does 2027 would be the end ???? I remember hyperscaler cashflow at all time low
surely they cant doing the same for 2028 right ??
gaigalas 1 days ago [-]
The fact that consumer-level DDR4 is getting more expensive makes absolutely no market sense to me.
wayne-werwolf 1 days ago [-]
I am guessing there is some fluidity between these markets. If consumers get priced out of the DDR5 market, the consumer DDR4 price increases as well. It is the same with cars, if the price goes up for new cars, it usually increases for used car market as well. Same with energy, price goes up for oil, all things beeing equal, other energy carries will increase to some degree.
Not sure if that's what you've meant...
gaigalas 1 days ago [-]
That's it. Which is insane. We used to buy used DDR4 from obsolete servers in buckets before it was going to become e-waste. Now even that one is expensive.
AdamN 1 days ago [-]
The lines that would have made consumer-level stuff are being reduced in favor of the higher profit margin/demand products. That means more supply of those other products (but still not enough) and lower supply of the consumer-level stuff (driving up prices).
gaigalas 1 days ago [-]
Isn't DDR4 like, old and obsolete? From my knowledge, you can't just take a fab that produces DDR4 and make it produce DDR5 or higher. It's kind of a leap that requires replacing a lot of stuff.
I would expect that old DDR4 machinery to end up somewhere else, still chugging along. Maybe someone buys the old production and keeps chugging low quality DDR4 or something.
To me, it sounds like that didn't happened. Fabs that were producing legacy DDR4 are still producing it, just selling it higher and surfing on the consumer price increase of DDR5. As DDR5 became more expensive, people looked to build old AM4 PCs with the older memory, and the market just raised the price so that option would also cost more. IMHO DDR4 price increase has nothing to do with production.
wao0uuno 14 hours ago [-]
>Isn't DDR4 like, old and obsolete?
I built a new gaming PC with DDR4 two months ago and it's pretty great. For servers RAM speed usually matters even less.
zahlman 15 hours ago [-]
> Isn't DDR4 like, old and obsolete?
I'm posting this from a machine that uses DDR3 and is still perfectly capable of meeting my needs.
gaigalas 33 minutes ago [-]
My comment was in the context of fabrication and how that affects the price of DDR4.
DDR5 fabrication requires a sub 15mm process, much more complex mounting and lots more testing equipment. It means you can't just quickly convert a DDR4 fab to DDR5.
It was meant to explain that DDR4 price increase has nothing to do with supply chain lanes, and it's more of an artificial demand manipulation than anything else.
phire 1 days ago [-]
There is some overlap. The last DDR4 fabs were typically on a 16nm process, the first DDR5 fabs were also on a 16nm process.
There shouldn't be much issue converting those 16nm DDR4 fabs to DDR5, it's really just an issue of swapping out the masks. That's assuming they haven't all already been converted years ago.
Even when there isn't an overlap, moving to the next process node usually isn't so much about replacing machines, but adding more machines. To oversimplify, you might not be able to convert one older DDR4 fab to DDR5, but sometimes you might be able to convert two older DDR4 fabs to one DDR5 fab.
Such conversions might not be the most cost-effective option in the long run (that unmodified fab could have kept selling DDR4 for years), but conversions might be significantly faster than building new fabs.
AdamN 1 days ago [-]
[not an expert so I could be totally wrong] I was also just talking about the physical space. Even if the machines on the entire line have to be replaced it still probably makes sense for a company to transition in situ - displacing the old line. There's the lease for the land and building, people's commutes are structured around a certain physical location, air conditioning and electricity are already there, etc... It's not just the big expensive machines that make up the fabrication - it's also the entire structure that makes that production happen.
phire 15 hours ago [-]
Even if they need to replace every single lithography machine, there are a lot of other machines that probably don't need to be replaced for any given upgrade.
Like everything related to it being a clean room; All the logistics robots which move wafers around; All the machines that deposit layers of material like CVD/AVD/Sputtering; The various machines that do the actual etching; The furnaces; The inspection and quality control equipment;
Also, absolutely everything to do with dicing and packaging. But that's usually already at another factory (often a completely different company) due to how process agnostic it is. Same applies to growing and prepping the silicon wafer.
gaigalas 23 minutes ago [-]
That is somehow inaccurate. DDR5 has significantly more challenging requirements.
The inspection and quality control equipment is one of those that you can't just reuse without careful transitioning. DDR5 has much higher speeds, on-die ECC and all sorts of complications DDR4 doesn't have.
15 hours ago [-]
sophrosyne42 1 days ago [-]
Because the prices of ddr5 are causing people to use substitutes
amelius 1 days ago [-]
If memory makers can fill their order books with only a single type of memory (whatever that is), then that would be preferable to them, I suppose.
kmfrk 1 days ago [-]
[dead]
TZubiri 14 hours ago [-]
So.. memory futures? Sounds like a signal of an efficient and mature market to me, manufacturers sell their stock way beforehand, keep production steady, and wholesalers can hold stock and absorb risk in exchange of profit.
Nothing prevents wholesalers from selling their memory or memory futures at a loss, as that is a risk that is justified by equal possibiltiy of the memory appreciating in excess of inflation.
ChiperSoft 8 hours ago [-]
I would not make that investment. All three memory makers are currently facing a US federal lawsuit for cartel price-fixing.
Why AI companies wont build their own RAM but prefer to pay multiple times more than what they sold pre AI boom? (And ram business was still profitable). They already make custom processors
johntb86 17 hours ago [-]
AI companies aren't fabbing processors. They're buying capacity at fabs to make their own designs. To make more RAM, they'd need to invest in making more DRAM fabs, which is a slow process.
tokioyoyo 17 hours ago [-]
Too slow, bum and bust cycles, dependencies one layer down are also booked up.
xg15 1 days ago [-]
Lame: Fight the war on general computation by spending decades pushing elaborate technical standards for locking down bootloaders, try to convince society why we need device attestation and hardware-bound keys, gradually establish devices without root access as the norm, get into constant skirmishes with free software and privacy advocates, get drawn into unpleasant political quagmires, etc etc
> As discussed previously, the ramp of HBM production will constrain industry supply growth in non-HBM products. Industrywide, HBM3E consumes approximately three times the wafer supply as D5 to produce a given number of bits in the same technology node. With increased performance and packaging complexity, across the industry, we expect this trade ratio for HBM4 to be even higher than the trade ratio for HBM3E. We anticipate strong HBM demand due to AI, combined with increasing silicon intensity of the HBM roadmap, to contribute to tight supply conditions for DRAM across all end markets. As the memory industry is still recovering from the challenging environment in 2023, this tight supply environment will help drive the considerable improvements in profitability and ROI (return on investment) that are needed to enable the investments required to support future growth.
https://investors.micron.com/static-files/4550f98c-1054-4847...
The problem is that they need actual chips tomorrow. I fear we are reaching the point in the semiconductor industry where, in order to sustain the revenue growth propping up their valuations, they're going to have to start selling future chips that cannot possibly be physically produced.
If there is demand for N chips at $X price, but you only have N/2 chips, then half the people aren't going to be able to buy them. The people who really need the chips, and would be willing to pay a lot more than $X for them, will be competing with people who are only willing to pay $X for them. You will end up getting scalping and shortages and hoarding. Since people know that there are enough people willing to pay a higher price, everyone will try to buy them at $X, even if they don't need them at all. Just buy them at $X, and immediately sell it to one of those companies willing to pay a lot more.
The market is extremely inefficient... since the manufacturer isn't charging enough, you get way too many people trying to buy them from the manufacturer.
So you find the price where the demand is N/2 chips, and everyone who is willing to pay that much will get one, and there is no profit for scalpers so the only people who will buy them will be actual companies that need them.
The people who really want the chips, and are able to pay a lot more than $X for them.
Please can y'all stop pretending that raising prices allocates resources by need and willingness to pay? With the level of spending power imbalance, the person who wants something and has too much money will ALWAYS squeeze out the people who actually need something and have less money. Always. No other factors are considered by the market.
That’s what makes this so distinct so it probably shouldn’t just be some offhand parenthetical at the end. It’s incredibly expensive, it takes a long time, and it’s incredibly specialized. You can absolutely forget the idea of a “disruptive startup” entering that space too.
Previously investing in that space didn’t make a lot of sense. Now that there are a lot of consumers willing to buy something they can build and sell at a price between the previous low point and the current high point.
The only tricky thing is that the current shortages need to continue until the fans are well up and running.
Even if the supply side suddenly crashes due to the AI bubble bursting or someone inventing a way to only need 1/10 of RAM, I can imagine them keeping the prices relatively high.
With the basic supply/demand curve, "shortage" and "discount" are synonyms and I know which word makes more sense to describe the scenario.
I don’t believe scalpers are an issue in B2B; these aren’t concert tickets being sold to the general public.
It’s clear you grasp the economic theory as it’s taught, but my comment was meant for you to question it. If someone is hungry you can charge more for food. The market allows and encourages it. But don’t mistake that for “willingness” and don’t mistake raising the prices purely to get extra money from the exchange as some inevitable law of the universe. You’re welcome to love the concept, but don’t whitewash it.
Demand is elastic. If the price of potatoes rises to $1000 / kg most people will switch to using pasta or rice for their dinner instead. There might be a shortage of potatoes at $1 / kg, but not at $1000 / kg.
If my 7-year-old computer were to die tomorrow I would've normally replace the entire thing. It has had a good life, the newer generations of hardware are a decent bunch faster, and it would be nice to get some additional features. With current RAM prices? No way I can afford to upgrade to DDR5, I'll have to get a replacement AM4 motherboard / CPU to fix it.
> I don’t believe scalpers are an issue in B2B
They are called "speculators". DRAM is a commodity and is traded no different from, say, potatoes. It's why there are companies like DRAMeXchange. DRAM module assemblers like Kingston, G.Skill, and Corsair will buy chips from whoever gives them the best deal. Similarly, anyone assembling hardware using DRAM will have stockpiled it when the AI boom became noticeable - with some almost certainly selling it off now that it has become too valuable to use for cheaper products.
The year 1845 called, and they'd like a word
If potatoes suddenly became $1000/kg, everyone would agree that there is a shortage of potatoes. While sure, people won't starve and will eat other things, it doesn't change the basic fact that there aren't enough potatoes.
The situation with RAM is even worse. If I'm trying to build and sell computing devices, I need RAM. Whether no one is willing to sell RAM at all, or whether they'll sell it only for 10x yesteryears's price, it doesn't make much difference to me, I'm still going out of business. Sure, that will then make demand technically go down, but not the shortage.
1) Some people won't be able to afford alternatives. They will starve to death
2) There may not be enough of other foods, so even if people have money, people will starve to death.
There are years where corn or rice or oats are of poor yield. A couple of years of that, and canned food and stored food is gone. It doesn't take long for "we have no <core staple food>", and things become bad.
This isn't a 3 month or year long blip in RAM pricing. If nothing changes, no bubble bursts, we won't see RAM become plentiful until 2035 at least. And a tight supply until 2030.
That's not a year of no potatoes. That's a decade of a core crop, with loads of calories, gone. That's "people are going to starve somewhere" trouble.
(I hate this food analogy, because in the West we've not seen a drought cause people to starve for half a century, if not longer. People are too complacent.
To speak to that, we grow crops where we get best value for the money. And best results. Corn cannot be grown in much of Canada, it needs a LOT of sun to mature, where as wheat can with a shorter growing season. Rice needs lots of water, where as other crops are more tolerant. Potatoes do well in certain soils. If there are no potatoes, you cannot just grow anything in that location.
Take away a core crop, and you can just lose those calories. Especially as farmers won't switch on the first year of a lost crop.)
RAM is quite durable, so I can imagine buying new servers with new CPUs, but instead of DDR5, it has DDR4 RAM slots. A loss of performance to be sure, but a good way to renew your fleet, and reduce the impact of all of this. But such boards aren't really here I don't think, not in quantity.
My MacBook is showing its age, though. Hearing that 2027 wont be any better makes me hopeful all the vibe coders can ask for memory efficient implementations to stretch existing hardware a little further.
In a lottery, the goods are misallocated to a bunch of people who aren't getting the most value out of the goods, creating economic inefficiency. A bunch of GPUs would be sitting in warehouses waiting to be resold (either when the price goes up, or in the inevitable black market), or in my basement screwing around with them, instead of being deployed in a way that the most people can benefit from creating maximum economic value from their deployment.
Your gaming PC isn't as economically valuable as JPMorgan using AI for fraud detection, for example.
Moreover, if you are appointed god and force everyone to sell GPUs at one dollar just because you want cheap GPUs, then this is the last batch of GPUs that will ever be produced and you'll have a shortage until the end ot time.
There are many other ways to allocate capacity. For example, you could give preferential access to customers who sign long-term purchase commitments, or some other favourable terms. These other methods might correlate well with value to the seller and might be easier for buyers to pay.
Other methods could be "allocate to customers that you want to maintain a good relationship with." I suspect HBM is not allocated either with pure auctions or by lottery.
If there was an Ozempic shortage tomorrow, it's not the morbidly obese people who "really need" it who would be getting it, it would be the Hollywood A-listers who need to drop 4 pounds for their next movie, or the billionaire who's feeling a little tubby. Price is a good allocation system, but it's important to recognize that it's not a mechanism for attaining some kind of ethically "best" allocation.
Sometimes this works and sometimes it does not.
If you happily sell apples for $1 and see a hungry person walking towards you, must you raise the price?
I think that basic thought gets lost sometimes when people talk about shortages causing the price to increase. The shortage didn’t cause anything, some executive decided they want more money. That’s all. There’s nothing inherent in the system that requires it. Whether that’s OK or not is up to the reader, I just think people lose sight of the reality and talk about it like it’s gravity, rather than simple decision making.
if you happily sell apples for $1 and now there are VC-funded companies who need a ton of apples to produce some dubious crap and can pay you $5 per. And you are the only company selling apples.
what you should do is keep selling $1 apples to retail customers and charge corporations as much as they accept. you invest it into more apple trees and maybe subsidize the hungry people even
> Now, you could raise prices to the point where you destroy demand.
You know supply and demand is like a curve right, you can find an optimal equilibrium? It’s not a cliff that you can fall off.
An airline overbooks flights. Lots of people get their flights cancelled. Takes six months for the fallout to settle where everyone who got burned booked their future flights on a different airline. 3 months out the airline has to cut the number of flights and cut prices in the face of falling revenue to reclaim market share. As soon as they do that, they're flooded with too many bookings. So they take the bookings and overbook flights again, but it takes time to bring the new flights online.
It's not a 1:1 situation. Each time you miss the market you wobble a little further until all the inefficiencies of bad predictions eat you up.
I’m confused because this reads like a denial of basic economics. If the price is high enough, the chip will be produced for you.
Do you mean because of the production lead time, higher prices won’t result in increased production? Commodities like corn have been managing this for a long time… what’s special about chips?
What is your actual argument?
Samsung, SK Hynix, and Micron have a combined market share of 90% - with most of the rest being a very new-to-the-market CXMT. It is a cutthroat market which behaves like a stereotypical "pork cycle". Semiconductor fabs cost billions to build and take years to complete, so you better be damn sure you have buyers before you start constructing one. You and your competition overestimated the demand? You have to pay back the construction cost, so you're now in a race to the bottom and one of you is going bankrupt.
Ever wondered where Intel came from? They started out as a DRAM manufacturer, which dominated their revenue well after the introduction of their first microprocessors. But in the early 1980s the glut of supply from new Japanese manufacturers made it so unprofitable that they had to ditch the memory market altogether. The stories of Texas Instruments and Motorola aren't much different. And that's not even mentioning the likes of Mostek, which once held a 85% market share and was dead less than 5 years later! Oh, and those Japanese manufacturers? All gone, pivoted like Intel or died like Mostek.
So no, the three remaining DRAM manufacturers aren't going behave like headless chickens and start ordering new fabs just because there's a bubble causing a temporary demand peak. Unless those AI companies are going to pay in advance, in cash, for an entire fab, they'll just have to wait and deal with the price increase.
So it's more "what's special about corn". It is also fairly hilarious to claim the parent is denying basic economics and then bring up corn as an example of having successfully managed economics. If the scales were not being thumbed, and "basic economics" were in play, corn would be in very very bad shape.
In the case of DRAM, there is an incredibly long history of these gloom/glut cycles, and they have stayed roughly the same timeframes (~3 years) since the 1990's.
Almost all the ones who have survived this long are either in the same kind of boat as corn - protected in various forms from the downside - or don't increase production and get caught out until they are absoultely forced.
The very temporarily increased profit is not worth going bankrupt for - they make more money long term by being very cautious and know this.
There are a near infinite number of economic studies you could look at (and several sibling comments cite some) - DRAM manufactuers don't chase the price and probably couldn't anymore if they want to.
None of this denies basic economic theory, of course, since economic theory is not exactly "rigorous", even to the degree it could be (IE even the parts that are pure analysis of data rarely reproduce!).
Memory is just too useful now that you can use it to drive cars and write code.
Which directly leads to the next big development: all the big players are investing in silicon with "baked-in" models, like [0,1]. Turns out you don't need an expensive general-purpose GPU with heaps of RAM to contain a model when you can make a custom ASIC around one specific model! Why spend a fortune on DRAM / HBM when all you need is some finetuning parameters which are easily stored in on-die SRAM?
[0]: https://www.theregister.com/systems/2026/08/06/amd-acquires-...
[1]: https://thenextweb.com/news/google-frozen-chip-gemini-silico...
Everyone. Open models are going to keep prices down. A lot of current models are more than usable. Self-hosting would have been an option if hardware prices get back to sane values.
The insane investments have to do with insane over-valuations, VCs involved and hence media spam on it. Chinese labs for example make do with 1% of the valuation and 1% of the resources.
Yesterday a coworker posted on a customer's Slack the specs of a box he is planning to buy to run local models. It's about 5k Euro. I am paying 18 Euro per month for Claude Pro and I'm going through a migration (almost a total rewrite) of a web app from Vue 2 / Vuetify 2 / Vuex to Vue 3 / Vuetify 4 / Pinia. I never hit the 6 hours limit. I could consider running an equivalent model on a 500 Euro machine (about 2 years of Claude Pro) but 5k is 20 years and that box will be obsolete or will have failed beyond repair (no spares) much earlier than that.
In any other business choosing (2) would mean someone else swoops in and steals all your business. It doesn't look like this is at all possible for memory fabs.
FWIW - Lots of people seem to believe the "in any other business", but it turns out there are tons and tons of places it is equally untrue, even where there is near zero time or cost barrier to entry.
You can use the state of the art from 20 years ago for that.
That's a classic stereotype of what an economic "expert" is going to say, while completely ignoring reality. Just because someone is willing to pay for it, does not mean, that goods just materialize. Sure, a lot of the time someone finds a way, but that is still limited by physical constraints: lack of expertise, lack of needed resources or as in this case simply lack of time.
"If the price is high enough" is of course technically true, but the scale of what high means in this context is important.
Chips aren't grown, for starters. They require very expensive facilities that take a long time to plan, build and start up. And ag firms are not seeing their stock price skyrocket based on the assumption that they will show exponential revenue growth every quarter, so there's no incentive for them to overbook their capacity and hope for the best. Maybe if the population was doubling every six months, you'd see something similar in corn. Prices in the real world are not perfectly elastic, and big deals are often locked in long before actual production.
It's a frictionless market in a vacuum right? No barriers to entry. You could make one right now if you wanted.
Maybe you could pay someone else who is less desperate to part with some, but that does not increase supply.
Nine women cannot produce a baby in a month, even though you could average about one per month if you wait about nine months.
They need it today. Best if it was NOW!
That's really interesting, and I wonder why? I believe HBM has redundant ECC bits by default, which would add a few %, but other than that, is it just that the yield is much lower due to die stacking? Of course, this is a 2024 document so things may have changed a bit since.
In order to get high bandwidth, you want memory as close as possible to the GPU. The more trace lane length = signal loss, bandwidth loss.. HBM is compact, and so you can stack 24GB modules, 8 around a GPU die.
If you tried to do that with normal memory, you need like 64 modules. So a a TON of traces more that all need to be equal length, and because so many = far away from the GPU = less bandwidth.
The issue is like stated above, its a process that waste a ton of wafers. Wafers that can make easily 3x more normal memory.
Intel with "Crescent Island" is trying to make a 160GB card using LPDDR5x memory but the bandwidth is only ~700GB/s.
The base interface die has eight or 12 memory dies stacked on top of it. The wafer cost of that interface die is therefore small.
If the process of die thinning and TSV stacking reduced yields by a factor of three, nobody would consider HBM mature enough to put into production, and especially not mature enough to be increasing stack height from one generation to the next.
Trace length has approximately nothing to do with die size. If anything, designing for shorter traces means you can get away with smaller PHYs at either end.
What might go some ways toward explaining such a huge difference in die size is that the TSVs themselves take up significant die area and must be fairly numerous to carry both a large number of signal wires and all the power and ground required by the stack. But it's wildly implausible that the die area consumed by the TSVs would be significantly larger than the die area consumed by the memory arrays themselves, or that anyone would build a memory die where the memory array was not a large majority of the total die area.
If there's any truth to that ~3x higher wafer requirement for the same number of bits as compared to DDR5, it must be a combination of several factors and probably includes something non-obvious and dubious, like counting all the area of the passive interposers that go between HBM stacks and GPUs (those interposers aren't competing for the same fab space).
What the freaking hell.
GPU is mostly fine and can probably reuse that but it’s Vega so not the greatest I was due for an upgrade but the prices of everything is ridiculous.
The thought of pc-of-Theseus’ing my busted ass PC is just sad.
Thank god I bought a PS5 Pro before Sony started jacking up the prices.
I remember when buying a $300 GPU was extravagant and gave you state of the art graphics.
Pick any B550 with a VRM heatsink for $90, a 5600 for $190, and you are back in business.
Get a 5070 or 9070 before GDDR supply is completely gone if you want an actual upgrade.
As for GPU price, that Vega 64 was $500 in 2018. Even a midrange GPU hasn't been $300 since the Radeon RX 480 launched a decade ago.
The PC I replaced when I bought this one was such a huge improvement.
My options basically boils down to fixing trash that’s falling apart or a sidegrade at best for the same price as what I bought a beast of a PC for 10 freaking years ago. Depressing.
I agree that the current situation sucks, but I don't think one can say it locks someone on your situation out of your game collection.
(edit: misremembered, $2000 is not the price of the game collection)
Point is, quite a lot of hardware can still do a lot. I have a Ryzen 5500U-based laptop that still outperforms the PCs of most people that I know (non-tech people).
Though in my case I feel like I have to get a loan to buy a workstation + an actual gaming PC for my wife. That all is beginning to look like at least $15k... yikes. But I am not sure I can wait until 2028 to see if things _start_ normalizing... Depressing indeed.
It wasn't the same model or vendor MB but it was a compatible one. Same chipset.
Same I just built a dual Broadwell Xeon E5 server / NAS. Why? Because someone local was selling a cheap Chinese X99 motherboard, two Xeons, and 64GB of DDR4 RAM for $150 CAD. The deal was too good. So I built a machine around it. (Though what I should have done was shove it into an old Dell T630/730, and saved myself some hassle.)
Now is the time of re-using old tech.
I've built two PCs in the last 5 years and both have had parts that were broken from the factory... There is something to be said about being able to hand off troubleshooting and replacement. I had ASRock pointing fingers at everything else to the point where I had swapped out CPU, PSU, and RAM, until they relented and replaced the motherboard (the replacement which is broken again now 5 years later by the way...)
That finally starts to ease pressure... then the AI RAM boom happens and now I can't even buy memory for anything I've got and I'm trapped in 8GB baseline RAM hell.
Every plan I had to update my network, my home, all out the window. I haven't seen my Steam library in at least a year or two now.
This isn't the condemnation of AI that you think it is.
When you are building an industry, you don't want to be in a "cash cow" phase. Because the person who is willing to reinvest profits (and further: seek and sink investments) into growth and innovation is going to eat your lunch.
~8 years ago my BIL condemned Tesla with "they aren't profitable" and I explained this to him. Now they're a trillion dollar company.
Uber tried that and failed. Now they cost as much as normal taxis if not more. Airbnb tried that and now cost as much as hotels if not more.
And tesla is a trillion dollar company on paper only. It can vanish into thin air in a day.
I also think most hardware expenditures done today will lose their value quickly; as we saw with crypto, hardware manufacturers will make specialized hardware so the cost/performance ratio will go down by at least 90% within a few years, possibly more.
It's going to look a bit like oligarchs subsidizing consumer tech as they try to offload more ram than God could own.
I hope that China makes up for the capacity that RAM producers are not willing to produce, so that the latter are sent back to the past (with no bailout nor protectionism)
I'm sure there will be a ton of standard DDR5 sticks, but not so much from AI-only datacenters.
Even better, looking forward to having hardware like that at home for "cheap" in the future, compared to what it costs right now :)
The average consumer isn't purchasing a lot of products with a lot of RAM every year.
Their 8GB of RAM phones will go up in price a little bit, but people aren't buying phones every year or even every other year.
So if the price of the 8GB of LPDDR went from $40 to $160 and it's all passed on to the consumer buying a new phone every 4 years, that's an extra $30/year in spending.
Most adults I know now use their phone for everything, so they're not buying new computers and laptops. They can wait 3 years for the market to settle before upgrading those, too.
Even I'm a heavy buyer of tech products and RAM, and I would bet you that my family's annual food bill fluctuates by more than what I've had to pay for RAM prices growing.
Many smaller companies will have issues getting enough, but even they could buy something with memory and rip it out. When you're willing to pay a hefty multiple, you can outbid other people. The shortages are not nearly bad enough to escape standard supply and demand.
>Most adults I know now use their phone for everything
You mean the cloud for everything. A lot of phone tasks share their computational and storage workloads off in the invisible ether that has actual computers with real costs behind them. Amazon has no problem with bumping up their costs in order to pay for their fleet of servers. I mean, what business doesn't use the cloud these days.
In the shorter term, you’ll probably see more “base” models on offer with some of the most intensive features disabled. There’s still going to be all of the telemetry stuff that makes them money through third parties though.
Total imports are maybe 14%? But not as highly tariffed.
Isn’t it just… prices?
https://en.wikipedia.org/wiki/Monetary_inflation
My solution is to see what four or five-year-old equipment I can buy that will let me run local LLMs. I may only get six or seven tokens per second out of an i7, but it's a start. And best of all, I can turn the machine off when I'm not using it.
IMO, Migrating to small-scale local LLMs would be a significant improvement over using data centers.
LLM serving is most efficient when you batch a lot of parallel requests together. Data center solutions also have the advantage of collecting queries from around the globe, so the hardware can be utilized around the clock.
Having everyone serve their own local LLMs would produce a lot more memory demand. Not less. The same memory would be idle most of the time, and when it was used it would be used for 1 person instead of a batch of requests.
There are other reasons to run local LLMs, but solving hardware demand problems is not one of them.
This shift is probably inevitable, but it will vary significantly by region depending on prices for electricity. Look at, for example, the difference in fundamental homelab build recommendations between Germans and just about anyone else. Electricity prices in Germany are so high that even a now expensive Raspberry Pi or other ARM board is often preferred over Intel/AMD builds due to low power draw (especially low idle power draw), an effect that adds up for a machine running all the time over years.
With local LLMs and the GPUs to run it, especially if you want a model available to you all the time and can remote into your local network to use it whenever you want, there's no escaping much higher power draws, even at idle. Wherever electricity is expensive, the electric bill can be a prohibitive barrier.
Eg: I shaved ~40W off the idle load on a server (250->210W) by doung nothing more than removing the redundant supply
A college buddy used to work at Motorola (I'm naming the company because they wouldn't mind this story being shared) back in the late 90's or early 2000's. They had redundant power to their campus, bought from two different companies, coming in on opposite sides of the campus, so that even if some backhoe operator cut a ground-based power line somewhere, they wouldn't lose power.
And yet, one morning, the power went off all across their campus. After a little investigation, they sent pretty much all their employees home at noon and told them "take the afternoon off, don't come back until tomorrow, you wouldn't be able to do any work anyway". Turns out that although the power lines came in at opposite sides of their campus, somewhere a few miles away both of the power lines feeding their campus ended up running through the same underground conduit. And yes, a backhoe had managed to cut that conduit and break both of the lines they depended on at the same time. They had a single, VERY non-obvious, point of failure, and the backhoe had unerringly homed in on that SPoF.
I'll second this. The combination of Whisper + LLM makes speech recognition fantastic. I occasionally have arm pain from typing, and this is a Godsend.
I don't use it to write code - but in my experience stuff like emails + docs was the greater source of pain (one generally types slower while coding).
> how do I do this one piece of a query in MySQL, since I'm not a MySQL DBA
Honestly for most things, the actual documentation is the best resource. I needed to learn to stop looking for solutions to my issue on various fora (like SO) and just read the primary documentation from the vendor. Have a look at dev.mysql.com or mariadb.com/docs/server
Most recent one I bought a multi-tb hard disk, but got an old multi-gb disk shoehorned into legitimate box. Couldn't just get a replacement - had to buy one for more money since that drive's price had increased.
https://pcpartpicker.com/trends/price/memory/
we would collectively all be better off by being forced to stick to older devices, value software optimized for proper resource management, and learn that instead of reflexively accepting "problems" we can see them as choices
enjoy the rat race
Reason being, a more powerful one would cost more than what I'm willing to pay (and, mostly, for fun and learning, which is priceless)
I have a minion in a terminal grinding away optimizing CPU-only LLM inference on Dual Xeon E5s. Why? Because it's neat.
How is a new business going to compete with incumbents that buy 100% of the necessary components needed to enter the market? Is no one else allowed to compete in this market anymore going forward?
HP, Acer and Asus are now using them
https://asia.nikkei.com/business/china-tech/hp-asus-and-acer...
Your computer already supports that. It's called "swap".
CPUs have L1, L2, L3+ caches too after all.
USB would be the limiter for both.
The M10 are over produced and only PCI-E 3.0 2x lane devices but 16GB of DDR5 is going to cost over $200.
You can do the same with USB-C M.2 adapters but the adapters cost more and eww USB for Swap, hope the device never drops.
https://www.amazon.com/GLOTRENDS-Adapter-Aluminum-Heatsink-P...
I suggest using an 1x PCI-e adapter because putting a M10 in a M.2 slot is likely going to only have 2 of 4 PCI-e lanes in use. You have to figure out the tradeoff of using a Swap on your, now very expensive as well, Flash storage that may have flashier sequential read and write speeds but doesn't have the ability to sustain random read and write performance which is what a Swap is going to use the most.
There are faster Optane drives that use 4 PCI-e lanes or even PCI-e 4.0, but they get pricey quickly, so you should just buy ram then. The deal is the 16GB Optane on an unused 1x PCI-e slot.
In any case DDR3 isn’t nearly as bad price-wise and would have much better energy efficiency than stacking 1–2 gig sticks.
For starters, The slowest sticks of DDR(x) are often slower than DDR(x-1). The issue is never capacity, but rather, performance.
The real nonsense? USB? It's a mess. Pick a USB cable and buy it from ANY retailer, let's make it easier, buy a USB-C cable. What is the data rate (depends on cable quality and length), Does it support power delivery? If so, how many watts? (~5W requires a very different cable from ~230W), how do you know from simply looking at the cable? If you buy a cable, how can you tell what it supports by simply looking at the connector? Imagine having a box full of USB-C cables. Could you tell me how fast each of those cables are? (The cables themselves don't! Many don't have any markngs at all, and if they do, the markings could be fraudulent)
RAM does not have that issue. A stick fits or it doesn't. Sure, there are a huge range of speeds, however, that range has a ballpark (JEDEC defines the ballpark, the "cartels" make the memory and push out some faster stuff).
While there are definite exceptions (I was bitten by one recently), you can generally plug in a DDR5 DIMM and expect it to work in the system.
The same cannot be said for USB. Some USB cables ONLY deliver power. Some only work with certain devices. I've USB-C (!!!) cables that only work with the devices they are shipped with, and even more annoyingly, the both may be true! ASUS (!!!) ships MOTHERBOARDS that can only pair with vital hardware via very specific USB-C cables (Strix Hive) and "GLORIOUS" ain't so glorious. Their mouses warn you not to use other USB-C cables, and they are right, depending on the make/model/generation, you can brick your "glorious" peripheral.
No, USB nonsense needs to stay far away from anything else.
Also, the reason this is a huge issue is because the memory makers are cartels. Only a few of them exist, they gang up and bully EVERYONE and nobody has ever invested money to create a competitor due to this, except China, which of course means that the U.S. and portions of the E.U. are insta-banning/trying to insta-ban, even though it is really freaking hard to install spyware on a memory module.
Okay, let's assume it's a USB-C cable and not just something USB-C shaped that pretends to be one.
> Does it support power delivery?
Yes.
> If so, how many watts?
Always at least 60W (3A). Up to 240W if it's e-marked.
> What is the data rate
Depends whether it's a USB 2.0-only cable (480Mbps) or a complete cable (20Gbps). Possibly more in Thunderbolt or USB4 modes - cables that handle those will be marked (both visually and with an e-mark). In any case, pretty easy to check just by plugging things in.
The last time I bought one I looked for one with a proper certificate but that eliminated the vast majority of cables on Amazon and I suspect that I ended up buying something ridiculously over engineered.
USB-C in C-to-C configurations require all the PD pins working for voltages to be supplied at all. So all USB-C cables that aren't the ones that actually don't work at all should be able to do that 20V x 3A -> 60W. Only in rare cases where a C-C cable is wired to make it look like one by being internally wired as C-A-A-C it becomes problematic since only USB-C has the config pins used for non-5V modes, but that should be actually rare.
I think USB-C shaped thing mentioned in GP refers to USB-C devices, like dollar store Bluetooth speakers and such. For those things, it's in fact not so rare that they come with tons of stupid quirks.
The can of worms is mainly in USB 3.0, Alt Mode, 5A mode, and Thunderbolt/USB4. 3A mode is fine. Raspberry Pi notoriously used a "5V 5A" mode that it's not supposed to be using.
If we're counting fake products then we need to count those RGB sticks that have no RAM in them.
Pretty sure those devs were saying the same thing way before 2017 as well, which seems to be ~ the last time RAM was this expensive.
Now RAM in the cloud, now you're really paying the Java premium.
Since that peak in 2014 until today, the prices were lower than now. The peak in 2017 was more than 10% lower than today.
To reach permanently higher prices than today, we must go backwards until 2011.
So we have already regressed at least 12 years into the past, but more likely 15 years, and it is unknown how much more we will regress.
It seems worth noting here that DDR5 RAM is faster than DDR3 RAM, and was not even available until 2020.
1. Make extra greasy pizza. Nauseating-level of grease.
2. Put pizza through De-Greasinator 5000.
Sounds efficient.
Were devs supposed to optimize RAM for a shortage that might come? I’m guessing you had enough foresight to stockpile RAM?
Every native UI framework I've ever used allows drawing directly to a canvas, and can have literally any visual style that can be created with pixels.
1) .net app, one text field, one button; private bytes 22mb, working set 27mb
2) native app, two text fields, three buttons; private bytes 1.2mb, working set 7.3mb
Just tested
Posting instead of researching in hopes someone smarter can chime in, because I'm lazy.
Also, you have Meta and Google investing in simplified performant versions of their stack for developing countries which is similar.
What's funny though, I see people that sometimes says it's cheaper to build the app they need in a single prompt than to search for it.
Moore's law is alive and well.
(PS: It is time to let ASML sell to CXMT.)
Other manufactures like Micron do have aggressive expansion plans, but fabs take a long time to build and bring online - multiple years.
https://www.guru3d.com/story/cxmt-dram-capacity-reportedly-s...
Traded as CUSU for end-users: https://en.twsc.com.cn/enCUSU/index.html
Nope!
https://www.reuters.com/world/asia-pacific/nanya-technology-...
https://www.blocksandfiles.com/flash/2026/03/27/three-ssd-ma...
https://thenextweb.com/news/nanya-6-billion-2027-ai-memory
The big three are also expanding their capacity. IMO, it seems to make more sense to sell equipments to those who could make better, efficient use of them. ASML has a backlog of €38.8 billion and is fully booked for 2027.
Why? High demand and low supply is very good for shareholders and make indexes go up, their goal is not to meet demand, they want exponential qurartly revunes
To me it seems like they’re just riding the AI hype out, expecting a crash after, and not willing to risk a lot of capital on someone else’s bubble.
If memory is now upstream of so many companies, then maybe a higher regulatory involvement would be appropriate (like power companies).
Specifically, and it could fix all of this, capping the max % of their supply they're allowed to sell into the AI space.
The arguments for infrastructure being managed more tightly by the government is because it's a natural monopoly that everything depends on. Memory isn't a natural monopoly, if we found ourselves in a spot where we all depend on 3 companies then that's not a law of nature, we need more companies.
Couldn’t intel repurpose some of its fabs (maybe the old ones?) to produce memory?
I’m no expert but… seems like an easy win?
Or lease a laptop from Apple if faceputer's aren't as much your style.
"The original retail price of the computer with 4 KiB of RAM was US$1,298 (equivalent to $6,900 in 2025)[21] and with the maximum 48 KiB of RAM, it was US$2,638 (equivalent to $14,020 in 2025)"
For the skeptical ones: Somewhere people are already being cut off from state and financial institutions without proprietary software with bundled security certificates, applications are bound to collect data on the environment they are used in (other apps) and organizations deliberately limit access to their services or cripple them without their applications (can't do things from a generic web-browser).
The words after this point make you less convincing, not more, as they only serve to confirm the egregious slippery slope of your reasoning.
I had previously ordered from Aeonfly, but they cancelled my order--likely because they wanted to raise the price.
Do deliver and they don't get paid because the buyer purchased with money that doesn't exist.
Lose-lose situation!
What we got in 2008 was central banks (eg Fed and ECB) willfully collapsing nominal GDP in their economies. Have a look at the dot-com bust or Black Monday for comparison.
What on earth are you talking about?
The credit risk was a very real problem; Kaupthing, Anglo Irish, RBS, Lehman etc.
In 2008 they didn't. Instead they actively tightened monetary policy by eg introducing interest on excess reserves. The ECB even hiked interest rates.
Companies defaulting on debt doesn't need to bring down the economy.
For the US, you can also see how the construction sector had been winding down for years (eg as measured in construction employment) without an impact on overall unemployment. The crisis was entirely avoidable.
See https://www.cato-unbound.org/2009/09/14/scott-sumner/real-pr... for a bit more background.
Take what's currently happening to Flock cameras as an object lesson. People are so fed up with Flock that they're cutting their cameras down en masse, sometimes wiping them clean from entire jurisdictions. Tens of thousands of other people in those jurisdictions are cheering them on, even offering alibis for them before they've been caught. Law enforcement doesn't seem too keen to catch them either. You absolutely can be so hated that the law will not be enforced to protect your property.
If AI companies don't slow down and build some good will, AI data centres are probably at risk. They'd probably have been targeted before Flock were they not harder targets. A higher degree of cooperation and organization will be required to sabotage them, but it would be foolish to believe it won't eventually happen.
Monopolizing memory production for another year suggests that lessons have not been learned, the mad build-out will continue, and we're headed for some truly crazy stuff.
And if AI is putting people out of jobs and making everything expensive - or that's just the perception - that's gonna set AI back for years.
It seems inevitable that these lost cameras will simply be upgraded to drone fleets. Not an identical replacement, but still "good enough" for the original purposes. Had they anticipated this problem and skipped the first phase, it might have been harder to manufacture consent for privacy invasion. But now anybody who was okay with the original cameras feels like a victim having them torn down, which makes it likelier they'll be okay with drones everywhere.
> Tens of thousands of other people in those jurisdictions are cheering them on, even offering alibis for them before they've been caught. Law enforcement doesn't seem too keen to catch them either. You absolutely can be so hated that the law will not be enforced to protect your property.
At least one third of the country is still okay with them, because they believe it helps rid them of illegals and deter crime, with downsides small enough to ignore, in their mind.
In sum, we don't have as much control over these things as you seem to believe. Capital was always in control with its crude but effective propaganda pipelines to manufacture enough consent. Now imagine those pipelines becoming even more effective thanks to AI. Not looking good.
Who cares about those people? I would argue even they themselves hardly do. And sure, they'll be even more okay with even worse stuff, and that even worse stuff will be even more hated by even more people, whatever the people who are "okay with" (that can mean ignorance, obedience, or evil intent, but it's not justification). If they can be okay with mass surveillance they can be okay with other things, so they're basically a wash, they're just okay with stuff.
> At least one third of the country is still okay with them
Then they can be okay with the majority saying you know what, we're sick of you, you can either stop having your toys or you will be segregated from us and live under your own surveillance. "Being okay with something" isn't some magical wand or a fortress, it's just a limp shrug. It means "pass".
> In sum, we don't have as much control over these things as you seem to believe.
I don't see the connection with the sentence before that, which is a minority being "okay with" something and a majority willing to fight and out there doing it. It's not about controlling what a minority of people are okay with, it's about changing what is done. The people who are okay with stuff can go read a comic book until it's over, how does that not solve anything you raised about them? Or are you saying their being okay with stuff should be respected or even heeded? Couldn't they just make that claim themselves, if they wanted to?
Imagine a raging house fire, people stumbling over each other trying to help, and some guy steps up and goes "WAIT!" they all look at him, sure he has a major contribution for interrupting something like this, a good idea, a plan perhaps; and he goes "... I'm actually okay with this."
It's not the majority, though. Roughly speaking, among adults, we have:
* 1/3 who are willing to vote for the red flavor of lesser-two-evils (some of them don't even register it as evil)
* 1/3 who are willing to vote for the blue flavor of lesser-two-evils (some of them don't even register it as evil, including when it literally manufactures consent for the first third to win next time around)
* 1/3 who see the above as a uniparty of capital interests offering the illusion of democracy through "close" elections as the bullshit that it is.
I wouldn't do that unless you want your airspace filled with birdshot
Civil disobedience? In this economy?
Nobody wants this robber baron shit
https://www.datacentermap.com/usa/michigan/ann-arbor/related...
It's all fair game IMHO, except the elephant in the room which is that the AI company bubble might pop.
It is also a term use in general trading anywhere traders open and close their positions quickly to make profit off an upswing which might be in part caused by traders doing this. If done with inside knowledge it is illegal, in this case more commonly referred to as “front running” (which is what many suspect happened with numerous ahemfortuitous/serendipitous position changes around announcements of changes in the state of the US vs Iran situation).
> Also, these companies supposedly intend to actually use the RAM they ordered, not reselling at some inflated price.
As well as the companies ordering for the purpose of DC roll-outs and upgrades, there will definitely be some buying purely to sell at a higher price a short time later.
> It's all fair game IMHO, except the elephant in the room which is that the AI company bubble might pop.
[and with reference to the sibling reply to this: “Then we'll have a lot of cheap RAM.”]
I don't think it will pop with a sudden bang, and even if it is it won't affect these memory sales or probably those for 2028. The first real effect will be those using the AI tools suddenly finding themselves needing to pay a lot more for them as companies offering those services deal with reduced access to “new” credit from the circling investment pool shore up their fanciful accounting that way instead. The DC builds/upgrades that are currently planned will happen but more funded by end users and less by the investment cycle. New plans might not be actioned at the same scale, but it will take a while for that to affect component prices as there is a lot of lead time involved.
A year ago the SOTA was GPT-5/Opus 4.1/Gemini 2.5 Pro, two years ago Sonnet 3.5.
Looking at Opus 5 for SOTA performance and GPT 5.6 Luna for a viable cheap alternative, AI is much more capable now.
Honorable mention: GPT 5.6 Sol on Cerebras, capacity limited to a few customers, is supposedly serving 750 tok/s.
Compared to 90 tok/s for non fast mode 5.6 Sol, 56 tok/s for Opus 5, and 190 tok/s for 5.6 Luna.
I am very curious about the next generation of models, GPT-6/Astra is rumored to launch still in August. Not sure what is the state of Anthropic's next Fable checkpoint.
If these models also deliver significant improvements, I really do not see how one could seriously still argue among the lines of AI being a scam, and the demand not being there to support the size of the investments.
I don’t think most people are saying it is purely a scam (well, some do but I don’t think they are to be taken seriously). What all these talks about circular financing and VC money are saying is that demand cannot sustain the sector long-term, not that there is no demand. People are certainly happy to pay say $20/month for whatever AI chatbot, but would they still pay if it were $200/month, which is closer to the actual costs.
There are several other details that point towards a unsustainable projections:
- measurable benefits from AI-ifying companies are nowhere near what is commonly believed. AI providers are hoping that they can keep the show going until the models are good enough, essentially faking it until they’ve made it, but that is not a given. It is also unstable because it is susceptible to change in public opinion.
- permits for new datacenters are not going to become easier to get as public opinion keeps turning against them. They will have to concentrate in friendly regions, which will add cost (more demand for the same location, plus interconnection for network and electricity, both of which can easily become bottlenecks).
- the electric grids are not ready for all those planned datacenters, so something will have to give. Building more and more on-site diesel generator in times where oil supply is so constrained and random is not very sustainable either.
- eventually the loans will come due and if earnings do not match there will be a, possibly severe, correction.
If we look at the historical example of the dot-com bubble, the crash was not caused by no demand. It was just caused by over-estimated demand and too much money going to a single sector of the economy. We still use the Internet, and it is still hugely important, but the correction was still severe and real people lost real money.
- Chatbot usage in $20 plans is likely nowhere close to a $200 cost.
Inference cost has come down rapidly, with reports from July claiming OpenAI can now serve all of the logged out ChatGPT traffic on just a few hundred GPUs.
- Measurable benefit: I doubt anything of value is being measured. MS Copilot with GPT 5.5 Instant processing SharePoint files? Developers using AI as a fancy autocomplete under a "I review every line" regime?
I rely on my personal value judgement, based on 30h/week I spend using AI outside of my regular job.
I am developing a mobile app, competing with companies with millions in revenue and entire dev teams. I know it is viable. Others lag in effective adoption, their opinion is likely to change soon with even more capable models.
- data center projects in the US: They look to me to mostly be constructed in the most remote backwater. If even there projects with such moderate environmental impact cannot be realized, that would be an embarrassing policy failure
- the grid: I think that one is true. AFAIK the constraint would be gas turbines, not diesel, and oil supply is not structurally constrained
- the loans: Anthropic is rumored to have become profitable earlier this year because of large growth in enterprise revenue. It does not look so bad to me
To add to that, as an ex-software dev that mostly works in a semi-unrelated field now (and can only code as a small part of my job): I think a lot of software devs are underestimating what someone with reasonable technical skills and specialised domain knowledge is able to create with AI.
I honestly wouldn't be surprised if, in 5 years' time, the majority of software used by (e.g.) potato farmers was primarily created by other potato farmers. The code might still be less elegant but I think it will be easier for the potato farmer to iterate with an AI than outsourcing to a dev firm.
I have little doubt that a competent potato farmer could possibly implement and sell such software today using Opus 5, Fable, or 5.6 Sol.
Now what I am going to say next may sound a bit crazy, and it is outside of my area of expertise. I hear the recent mathematical breakthroughs made by GPT-6/Astra are field medal worthy discoveries.
What if the current trajectory of improvement holds for another year or two?
Is it impossible that LLMs gain superhuman ability to reason over a large number of constraints so that they can make novel breakthroughs unimaginable to us today?
What I hope to see in five years is not potato farmers writing software, but programmable immune cells that safely kill cancer.
Perhaps I am an optimist.
AI grows and grows and becomes god and kills us all.
How did it all start? Oh we invested all our money in AI.
We rewrote all our Python code in Go because the RAMs too damn high!
Part of me also thinks we need to stop this. How can we have a democratic society when only billionaires have access to computing equipment?
Besides the AI craze, to me it sounds kinda reasonable. Most large buyers (apple, public clouds, OEMs) do forecasting and orders in advance, right?
surely they cant doing the same for 2028 right ??
Not sure if that's what you've meant...
I would expect that old DDR4 machinery to end up somewhere else, still chugging along. Maybe someone buys the old production and keeps chugging low quality DDR4 or something.
To me, it sounds like that didn't happened. Fabs that were producing legacy DDR4 are still producing it, just selling it higher and surfing on the consumer price increase of DDR5. As DDR5 became more expensive, people looked to build old AM4 PCs with the older memory, and the market just raised the price so that option would also cost more. IMHO DDR4 price increase has nothing to do with production.
I built a new gaming PC with DDR4 two months ago and it's pretty great. For servers RAM speed usually matters even less.
I'm posting this from a machine that uses DDR3 and is still perfectly capable of meeting my needs.
DDR5 fabrication requires a sub 15mm process, much more complex mounting and lots more testing equipment. It means you can't just quickly convert a DDR4 fab to DDR5.
It was meant to explain that DDR4 price increase has nothing to do with supply chain lanes, and it's more of an artificial demand manipulation than anything else.
There shouldn't be much issue converting those 16nm DDR4 fabs to DDR5, it's really just an issue of swapping out the masks. That's assuming they haven't all already been converted years ago.
Even when there isn't an overlap, moving to the next process node usually isn't so much about replacing machines, but adding more machines. To oversimplify, you might not be able to convert one older DDR4 fab to DDR5, but sometimes you might be able to convert two older DDR4 fabs to one DDR5 fab.
Such conversions might not be the most cost-effective option in the long run (that unmodified fab could have kept selling DDR4 for years), but conversions might be significantly faster than building new fabs.
Like everything related to it being a clean room; All the logistics robots which move wafers around; All the machines that deposit layers of material like CVD/AVD/Sputtering; The various machines that do the actual etching; The furnaces; The inspection and quality control equipment;
Also, absolutely everything to do with dicing and packaging. But that's usually already at another factory (often a completely different company) due to how process agnostic it is. Same applies to growing and prepping the silicon wafer.
The inspection and quality control equipment is one of those that you can't just reuse without careful transitioning. DDR5 has much higher speeds, on-die ECC and all sorts of complications DDR4 doesn't have.
Nothing prevents wholesalers from selling their memory or memory futures at a loss, as that is a risk that is justified by equal possibiltiy of the memory appreciating in excess of inflation.
https://www.polygon.com/ram-manufacturers-sued-supply-price-...
SK Hynix's suppliers are also under investigation for anti-trust behavior in South Korea.
https://en.sedaily.com/society/2026/05/29/prosecutors-raid-s...
Game: Just buy up all the RAM...