If LLMs' utility continues to scale with size (which seems likely as we begin training embodied AI on a massive influx of robotic sensor data) then it will continue to gobble up memory for the near future. We may need both increased production capacity _and_ a period of more efficient software development techniques as was the case when a new 512kb upgrade cost $1,000.
Most DRAM is already purchased through contracts with manufacturers.
Manufacturers don't actually want too many extremely long term contracts because it would limit their ability to respond to market price changes.
Like most commodities, the price you see on places like Newegg follows the "spot price", meaning the price to purchase DRAM for shipment immediately. The big players don't buy their RAM through these channels, they arrange contracts with manufacturers.
The contracts with manufacturers will see higher prices in the future, but they're playing the long game and will try to delay or smooth out purchasing to minimize exposure to this spike.
> Additionally, we're likely to see Chinese fab'd DRAM now, which they've been attempting since the '70s but never been competitive at.
Companies like Samsung and SK Hynix have DRAM fabs in China already. This has been true for decades. You may have Chinese fab'd DRAM in the computer you're using right now.
Are you referring to complete home-grown DRAM designs? That, too, was already in the works.
> Manufacturers don't actually want too many extremely long term contracts because it would limit their ability to respond to market price changes.
I don't agree with this sentence. Why would not the same apply advice to oil and gas contracts? If you look at the size and duration of oil and gas contracts for major energy importers, they often run 10 years or more. Some of the contracts in Japan and Korea are so large, that a heavy industrial / chemical customers will take an equity stake in the extraction site.Except silicon, power, and water (and a tiny amount of plastic/paper for packaging), what else does a fab need that only produces DRAM? If true, then power is far and away the most variable input cost.
Because oil & gas suppliers only ever sell one product, and memory fabs can dynamically switch product mix in response to supply & demand to optimize profits. The same sand, power and water can make DDR4, HBM or DDR5
Various chemicals too, https://haz-map.com/Processes/97
Yes, via cxmt as discussed by Asianometry here: https://www.youtube.com/watch?v=mt-eDtFqKvk
As I mentioned, various groups within China have been working on China-native DRAM since the '70s. What's new are the margins and market demand to allow them to be profitable with DRAM which is still several years behind the competition.
A Japanese factory that made epoxy resin for chips was destroyed and the price of SIMM chips skyrocketed (due to lack of availability).
I remember being very upset that I wasn't going to be able to upgrade to 4MB.
But yes we're going to need more fabs for sure
If the shortage of RAM is because of AI (so servers/data centers I presume?), wouldn't that mean the shortage should be localized to RDIMM rather than the much more common UDIMM that most gaming PCs use? But it seems to me like the pricing is going up more for UDIMM than RDIMM.
Wouldn't that mean that a shortage of DRAM chips should cause price difference in all of them? Not sure that'd explain why RDIMM prices aren't raising as sharply as UDIMM. That the fab and assembly lines have transitioned into making other stuff makes sense why'd there be a difference though, as bradfa mentioned in their reply.
The manufacturers make the individual chips, not the modules (DIMMs). (EDIT: Some companies that make chips may also have business units that sell DIMMS, to be pedantic.)
The R in RDIMM means register, aka buffer. It's a separate chip that buffers the signals between the memory chips and the controller.
Even ECC modules use regular memory chips, but with extra chips added for the ECC capacity.
It can be confusing. The key thing to remember is that the price is driven by the price of the chips. The companies that make DIMMs are buying chips in bulk and integrating them on to PCBs.
Quite a few unbuffered designs in the past had a "missing chip". If you ever wondered why a chip was missing on your stick, it's missing ECC. Don't know if it's still the case with DDR5 though.
Also with DDR5 each stick is actually 2 channels so you get 2 extra dies.
Was reading a series of displeased posts about it. Can't seem to find it now.
Also vary a bit between winter a summer, basically in winter they can get away with putting a bit more volatile compounds coz it's colder
I am hoping some of that Clayton Christensen disruption the tech theocracy keep preaching about comes along with some O(N) decrease in transformer/cDNN complexity that disrupts the massive server farms required for this AI boom/bubble thing.
Perfectly stated. I think comments like the one above come from a mentality that the individual consumer should be the center of the computing universe and big purchasers should be forced to live with the leftovers.
What's really happening is the big companies are doing R&D at incredible rates and we're getting huge benefits by drafting along as consumers. We wouldn't have incredible GPUs in our gaming systems and even cell phones if the primary market for these things was retail entertainment purchases that people make every 5 years.
Computing didn't take off until it shrank from the giant, unreliable beasts of machines owned by a small number of big corporations to the home computers of the 70s.
There's a lot more of us than them.
There's a gold rush market for GPUs and DRAM. It won't last forever, but while it does high volume sales at high margins will dominate supply. GPUs are still inflated from the crypto rush, too.
The iPhone isn't exactly a consumer computation device. From that perspective, it does less work at a higher cost.
(I remember the huge window in which phone companies desperately put out feature phones with sub-par touch screens, completely missing the value to consumers. The iPod Touch should've been warning enough... and should've been (one of) my signal(s) to buy Apple stock, I guess :-)
Arguably we don't. Most of the improvements these days seem to be on the GPGPU side with very little gains in raster performance this decade.
Games were always going to go 3d sooner or later, the real pressure of the high volume competitive market got us more and more capable chips until they were capable enough for the kind of computation needed for neural networks faster than a slow moving specialty market could have.
Yes. That is my point. The customers willing to pay the high initial R+D costs opened up the potential for wider adoption. This is always the case.
Even the gaming GPUs which have grown in popularity with consumers are derivatives of larger designs intended for research clusters, datacenters, aerospace, and military applications.
No question that chip companies are happy to take consumers money. But I struggle to think of an example of a new technology which was invented and marketed to consumers first.
A colleague who worked with me about 10 years ago on a VDI project ran some numbers and showed that if a Time Machine were available, we could have brought like 4 loaded MacBook Pros back and replaced a $1M HP 3PAR ssd array :)
Compute is cheaper than ever. The ceiling is just higher for what you can buy.
Yes, we have $2000 GPUs now. You don't have to buy it. You probably shouldn't buy it. Most people would be more than fine with the $200-400 models, honestly. Yet the fact that you could buy a $2000 GPU makes some people irrationally angry.
This is like the guy I know who complains that pickup trucks are unfairly priced because a Ford F-150 has an MSRP of $80,000. It doesn't matter how many times you point out that the $80K price tag only applies to the luxury flagship model, he anchors his idea of how much a pickup truck costs to the highest number he can see.
Computing is cheaper than ever. The power level is increasing rapidly, too. The massive AI investments and datacenter advancements are pulling hardware development forward at an incredible rate and we're winning across the board as consumers. You don't have to buy that top of the line GPU nor do you have to max out the RAM on your computer.
Some times I think people with this mentality would be happier if the top of the line GPU models were never released. If nVidia stopped at their mid-range cards and didn't offer anything more, the complaints would go away even though we're not actually better off with fewer options.
If the result was that games were made and optimised for mid-range cards, maybe regular folks actually would be better off.
-the whole reason why the GPU is $2000 is because of said AI bubble sucking up wafers at TSMC or elsewhere, with a soupçon of Jensen's perceived monopoly status...
-for a good part of the year, you could not actually buy said $2000 GPU (I assume you are referring to the 5090) also because of said AI bubble
(granted, while Jensen does not want to sell me his GPU, I would like to point out that Tim Cook has no problem taking my money).
on that point, I can go and buy a Ford F150 tomorrow. Apparently, per the article, I would have problems buying bog standard DDR5 DIMMS to build my computer.
The move to cloud computing and now AI mean that we're back in the mainframe days.
You still do. There is no "AI movement" you need to participate in. You can grab a copy of SICP and a banged up ten year old thinkpad and compute away, your brain will thank you. It's like when people complain that culture is unaffordable because the newest Marvel movie tickets cost 50 bucks, go to the library or standardebooks.org, the entire Western canon is free
Living on the edge from 4 years ago is basically free.
Hard disagree. A $600 Mac Mini with 16GB of RAM runs everything insanely faster than even my $5000 company-purchased developer laptops from 10 years ago. And yes, even when I run Slack, Visual Studio Code, Spotify, and a gazillion Chrome tabs.
The HN rhetoric about modern computing being slow is getting strangely disconnected from the real world. Cheap computers are super fast like they've never been before, even with modern software.
That an 11-year-old PC can keep up today (with or without an upgrade) is evidence that systems are keeping up with software bloat just fine. :)
Bullshit. It was cramped and I wasn't able to do half of what I was wanting to actually do. Maybe it was plenty for your usecases, but such a small amount of memory was weak for my needs in the late 90s and 2000s. 64MB desktops struggled to handle the photo manipulations I wanted to do with scanned images. Trying to do something like edit video on a home PC was near impossible with that limited amount of memory. I was so happy when we managed to get a 512MB machine a few years later, it made a lot of my home multimedia work a lot better.
The manufacturers are scumbags is more likely answer.
Anyway, that's the kind of market that governments always need to act upon and either supply directly or regulate intensively.
But my guess is that this shortage is short-lived (mostly because of the threat above). There's no OPEC for tech.
Maximizing profit is the only sane way to play a rigged game
Unforseen things like the pandemic hurt profits.
Letting things go this unmanaged with a 3 year run way for AI demand seems a little hard to understand. In this case, not anticipating demand seems to creates more profit.
Next is probably CPUs, even if AIs don't use them that much, manufactures will shift production to something more profitable, then gouge prices so that only enterprises will pay for them.
What's next? Electricity?
Where the f*k is all the abundance that AI was supposed to bring into the world? /rant
Things being too cheap allows money to pool at the bottom in little people's hands in the forms of things like "their homes" and "their computers" and "their cars".
You don't really want billions in computing hardware (say) being stashed down there in inefficient, illiquid physical form, you want it in a datacentre where it can be leveraged, traded, used as security, etc. If it has to be physically held down there, ideally it should be expensive, leased and have a short lifespan. The higher echelons seem apparently to think they can drive economic activity by cycling money at a higher level amongst themselves rather than looping in actual people.
This exact price jump seems largely like a shock rather then a slow squeeze, but I think seeing some kind of reversal of the unique 20th century "life gets better/cheaper/easier every generation".
To me the #1 most important factor in a maintaining a prosperous and modern society is common access to tools by the masses, and computing hardware is just the latest set of tools.
In the hands of the owners of the AI, as a direct consequence of the economic system. It was never going to play out any other way.
Humanity will have to adopt new human-focused modes of living and organizing society, or else. And climate change is coming along to make sure the owning class can't ignore this fact any longer.
AI will lead to abundance. For those that own stuff and no longer have to pay for other people to work for them.
Yes. My electricity prices jumped 50% in 3 years.
How much is due to long overdue infrastructure upgrades and greed by providers, vs the cost of energy?
Also, consumer prices _have_ risen (mine included), but it's not clear that this is only because AI. While EV charging is not at the scale of all data centers combined, it seems to grow even faster than the datacenter's consumption, and is expected to eclipse the latter around 2030. Maybe sooner due to missing solar incentives.
Also, to rant on: According to [1], an average Gemini query costs about 0.01 cents (Figure 2 - say 6000 queries per kWh at 60 cents/kWh, which is probably more than the industrial consumers pay). The same paper says one other providers is not off by that much. I dare say that at least for me, I definitely save a lot of time and effort with these queries than I'd traditionally have to (go to library, manually find sources on the web, etc), so arguably, responsibly used, AI is really quite environmentally friendly.
Finally: Large data centers and their load is actually a bit fungible, so they can be used to stabilize the grid, as described in [2].
I would think it would be best if there were more transparency on where the costs come from and how they can be externalized fairly. To give one instance, Tesla could easily [3] change their software to monitor global grid status and adjust charging rates. Did it happen ? Not that I know. That could have a huge effect on grid stability. With PowerShare, I understand that vehicles can also send energy back to power the house - hence, also offload the grid.
[1] https://services.google.com/fh/files/misc/measuring_the_envi...
[2] https://www.linkedin.com/feed/update/urn:li:activity:7358514...
[3] that's most likely a wild exaggeration
If you make $4k/mo and rent is $3k, it's pretty silly to state that it's a meaningful thing for someone to scrimp and invest $100/mo into a brokerage account.
They definitely should do this, but it's not going to have any meaningful impact on their life for decades at best. Save for a decade to get $12k in your brokerage account, say it doubles to $24k. If you then decide you can get a generous 5% withdrawal rate you are talking $600/yr against rent that is now probably $3500/mo or more. Plus you're killing your compounding.
It's good to have so emergencies don't sink you - but it's really an annoying talking point I hear a lot lately. Eye rolling when you are telling someone struggling this sort of thing.
It really only makes a major impact if you can dump large amounts of cash into an account early in life - or if you run into a windfall.
I remember when there was a flood somewhere in Thailand in the 2011 and the prices of hardisks went up through the roof.
https://www.forbes.com/sites/tomcoughlin/2011/10/17/thailand...
That can be a bigger problem for civilization.
next?
https://www.cnbc.com/2025/11/14/data-centers-are-concentrate...
What is your source on that? Moore's Law is Dead directly contradicts your claims by saying that OpenAI has purchased unfinished wafers to squeeze the market.
Tom’s Hardware: “Samsung raises memory chip prices by up to 60% since September as AI data-center buildout strangles supply,” Nov 2025. https://www.tomshardware.com/tech-industry/samsung-raises-me...
Note the consistent "up to 60% since September" figure in the above recent reports. That's for one module capacity, with others being up 30% to 50% - and it certainly isn't the 200% or more we're apparently seeing now in the retail market. That's pure panic hoarding, which is actually a very common overreaction to a sudden price spike.
That'll come with the bubble bursting and the mass sell off.
more money for shareholder, 5 Trillion Nvidia???? more like a quadrillion for nvidia market cap
It may have been a bit self-deprecating, but I think your “rant” is a more than justified question that really should be expanded well beyond just this matter. It’s related to a clear fraud that has been perpetrated upon the people of the western world in particular for many decades and generations now in many different ways. We have been told for decades and generations that “we have to plunder your money and debase of and give it to the rich that caused the {insert disaster caused by the ruling class} misery and we have to do it without any kind of consequences for the perpetrators and no, you don’t get any kind of ownership or investment and we have to do it now or the world will end”
Abundance isn't even the right framing. What most people actually want and need is a certain amount of resources - after which their needs are satiated and they move onto other endeavors. It's the elites that want abundance - i.e. infinite growth forever. The history of early agriculture is marked by hunter-gatherers outgrowing their natural limits, transitioning to farming, and then people figuring out that it's really fucking easy to just steal what others grow. Abundance came from making farmers overproduce to feed an unproductive elite. Subsistence farming gave way to farming practices that overtaxed the soil or risked crop failure.
The history of technology had, up until recently, bucked this trend. Computers got better and cheaper every 18 months because we had the time and money to exploit electricity and lithography to produce smaller computers that used less energy. This is abundance from innovation. The problem is, most people don't want abundance; the most gluttonous need for computational power can be satisfied with a $5000 gaming rig. So the tech industry has been dealing with declining demand, first with personal computers and then with smartphones.
AI fixes this problem, by being an endless demand for more and more compute with the economic returns to show for it. When AI people were talking about abundance, they were primarily telling their shareholders: We will build a machine that will make us kings of the new economy, and your equity shares will grant you seats in the new nobility. In this new economy, labor doesn't matter. We can automate away the entire working and middle classes, up to and including letting the new nobles hunt them down from helicopters for sport.
Ok, that's hyperbole. But assuming the AI bubble doesn't pop, I will agree that affordable CPUs are next on the chopping block. If that happens, modular / open computing is dead. The least restrictive computing environment normal people can afford will be a Macbook, solely because Apple has so much market power from iPhones that they can afford to keep the Mac around for vanity. We will get the dystopia RMS warned about, not from despotic control over computing, but from the fact that nobody will be able to afford to own their own computer anymore. Because abundance is very, very expensive.
The memory business is a pure commodity and brutally cyclic. Big profit => build a fab => wait 2 years => oh shit, everyone else did it => dump units at below cost. Repeat.
I'm now seeing $480 CAD for a single stick.
> Memory: 16 GB GDDR6 SDRAM
So unless the RAM price jumps to 4x the price of a PS5, getting a PS5 is not the most cost efficient way to get to 64 GB of RAM.
In comparison, PS3 has been used to build cheap clusters[2].
Big manufacturers also order their DRAM in advance with contractually negotiated pricing. They're not paying these spot market prices for every computer they ship.
~~~~~helps that Apple's SoC has the RAM on the main die itself. They're probably immune from these price hikes, but a lot of the PC/Windows vendors would, which would only make Apple's position even stronger~~~~
If Apple is insulated it is likely because Apple signs big contracts for large supply and manufacturers would prefer to be insulated from short-term demand shocks and have some reliability that their fabs can keep running and producing profitable chips.
And in this article you can see a photo of the memory chips attached outside of the Apple component https://www.gizmochina.com/2020/11/19/apple-mac-mini-teardow...
It used to be a general rule of thumb that you could build a computer of roughly equivalent power for the cost of a game console, or a little more — now the memory costs more than the whole console.
The PS5 has 16GB of RAM. One can buy 16GB of RAM for ~$100 [1].
[1] https://pcpartpicker.com/product/9fgFf7/kingston-fury-beast-...
But since it's 16 GB, the comparison doesn't really make sense.
That's a an analogy-- a literary technique the writer is using, to show the correspondence between the price of a specific amount of DDR5 RAM to a fully integrated system, so the reader can follow the conclusions of their article easier.
Gaming consoles are something people buy. Any parent or gamer has an idea what they cost.
People do not buy London Buses themself
I wonder what you'd think if bus tires exploded in price and started costing .25 London busses per tire.
I'd bet that a good chunk of the apparently sudden demand spike could be last month's Microsoft Windows 10 end-of-support finally happening, pushing companies and individuals to replace many years worth of older laptops and desktops all at once.
I got 2 sticks of 16GB DDR4 SODIMM for €65.98 back in February. The same two sticks in the same store now cost €186
I'm joking, but only kind of. It's not a domain that I do a lot of, but I haven't touched Qt in so long that it would basically be starting from scratch if I tried to write an app with it; I could write an Election app in like an hour.
Bad time if you need to build a computer.
Looking at it optimistically, you're probably going to find DDR5 at bargain prices again in 2027.
When do you think prices will recede again?
And because everything needs prices expect all electronics to be ~20%-80% more expensive in 2027 compared to today, naturally this includes the profit margin.
and naturally every regulation related companies don't like will supposedly be at fault for this (e.g. right to repair)
at least that is a wild speculation on my side
People just have to wait. As prices are sky high, production capacity will likely increase. Some AI companies will go bust. Demand will plummet and we will buy RAM for pennies while the market consolidates.
Commander Data's specifications in the Star Trek TNG episode The Measure of a Man from 1989: 800 quadrillion bits of storage, computing at 60 trillion operations per second.
100 petabytes. That's a big machine. A very big machine. But supercomputers now have memories measured in petabytes.
They never used "bits" again in any Star Trek script. It was kiloquads and gigaquads from then on.
Then I did some googling and it turns out that a single 5090 GPU has a peak FP32 performance of over 100 TFLOPS!
AI: RAM
Thanks for taking away years of affordable computing from people. Time is more valuable; there's no getting it back.
I have a gaming PC, it runs Linux because (speaking as a Microsoft sysadmin with 10 years under my belt) I hate what Windows has become, but on commodity hardware it’s not quite there for me. Thought I’d play the PlayStation backlog while I wait for the Steam Machine.
I guess I'm glad I bought when I did; didn't realize how good of a deal I was getting.
Felt like I overpaid at the time too. Wow
AI/LLM companies will pay TSMC more than Apple is willing to further subsidize this neat little box.
> Despite server-grade RDIMM memory and HBM being the main attractions for hardware manufacturers building AI servers, the entire memory industry, including DDR5, is being affected by price increases. The problem for consumers is that memory manufacturers are shifting production prioritization toward datacenter-focused memory types and producing less consumer-focused DDR5 memory as a result.
But I'm sure the hysteria around that isn't helping prices come back down either.
ebay .com /itm/ 256168320806
(no association, just trying to help, I am still using DDR4)