This is a most excellent place for technology news and articles.
AI increases my power utility bill
AI takes my water
AI increases the price of GPUs
AI increases the price of RAM
AI makes my search results worse and slower
AI is inserted into every website, app, program, and service making them all worse
All so businesses and companies can increase productivity, reduce staff, and then turn around and increase prices to customers.
Isn't capitalism a blast?
Right, what you said, but I'd remove the part where it increases productivity.
Seriously. The amount of time spent handholding the AIs no one asked to use more than offsets any supposed productivity gains.
It's a win-win with staff layoffs. Businesses that want to lay people off have a convenient scapegoat and AI companies receive undeserved praise.
A win-win for everyone but the employees, of course.
Increase productivity?
People buying RAM: oh no, what do we do?
People buying GPUs: first time?
What?
You think this is the first ram crunch?
It's not even the first one in this decade...
Well, I was aware RAM prices fluctuate.
I've never been so unfortunate when buying larger RAM, or building a new system with a new DDR version.
Those are the same people.
OpenAI’s “Stargate” project has recently signed an agreement with Samsung and SK hynix for up to 900,000 wafers of DRAM per month. That figure alone would account for close to 40% of global DRAM output.
High-density NAND products are effectively sold out months in advance. Samsung’s next-generation V9 NAND is already nearly booked before it's even launched. Micron has presold almost all of its High Bandwidth Memory (HBM) output through 2026. Contracts that once covered a quarter now span years, with hyperscalers buying directly at the source.
If China's going to compete on AI, it's going to be doing so with a limited supply of memory, I expect.
They're going all in on domestic chip manufacturing, and they are catching up much faster than armchair generals, and even actual generals, predicted.
Nice article but the numbers are a lot lower here in the EU.
While there is some pricing increase it's currently more around 50% and not 100%.
The selected kit is also extremely expensive (350€ was ~300€) - similar kits are available for a lot less (270€ was ~180€) - so I doubt that anyone was buying it in the first place.
I also think it's not completely AI related but more likely that this is another RAM price fixing scandal happening right now. Pretty much the same that we see today happend in 2017-2018.
God fucking damn it... Now they want our RAM?!
I need that for all my Chrome tabs and pet protogen, you fucking clankers!
Doesn't Windows 11 in practice require even more memory than Windows 10 to operate with decent performance?
Meanwhile my Linux gaming PC seems to actually use less memory than back when it was a Windows machine.
My work laptop was upgraded to Windows 11 and performance has severely suffered.
As someone who usually uses 3 monitors (sometimes 4) and does GIS, it's an issue.
This seems like an appropriate place for me to bitch:
2 months ago I bought a new pre-built pc. It should've had 64gb of ram but had 32gb. They said the sticks they used were out of stock so they gave me a credit for $100 USD. I spent the 100 on 32gb more of what I thought was the exact same ram. I fucked up and bought a slightly higher speed so they wouldn't work together after I tried for an afternoon. I also checked the correct listing i should've bought but it was more expensive, at about $125.
I gave up and decided I'd just buy the faster ram again when it came back, rather than return it and get the correct one. It went out of stock in the time it took me to get my order so I figured I'd just wait.
2 MONTHS later, it never came back in stock but an almost identical pair, with slightly different timing, is in stock right now at $216. If i had any idea this was coming in just 2 months, I could've just bought 64gb at once and started fresh, or corrected my mistake by returning what I bought.
So i guess I'll continue waiting, but hey at least notepad has copilot in it.
I always thought ram of different speeds worked together, they just were run at the speed of the slowest stick.
First crypto miners came for my video cards, then AI came for my DIMMS...
Next tech-sector grift will probably go for our network adapters or some shit..
Remember when "Chia coin" came for our HDDs too?
https://www.tomshardware.com/news/hard-drive-prices-skyrocket-asia-scalpers-making-bank
I just got a 2x64GB 6000 kit before its price skyrocketed by like $130. I saw other kits going up, but had no clue I timed it so well.
...Also, why does "AI" need so much CPU RAM?
In actual server deployments, pretty much all inference work is done in VRAM (read: HBM/GDDR); they could get by with almost no system RAM. And honestly most businesses are too dumb to train anything that extensively. ASICs that would use, say, LPDDR are super rare, and stuff like Hybrid/IGP inference is the realm of a few random folks with homelabs... Like me.
I think 'AI' might be an overly broad term for general server buildout.
Same memory production capacity can be allocated to ddr5 or to hbm and openai signed contracts with sk hynix and samsung, the two largest ram manufacturers in the world, and bought a significant percentage of next year's production.
DDR5 prices started spiking as that deals impact propagated through the supply chain. I bought a 2x32 6800 Cl30 kit for 195 euro 12 days ago. It was 330 euro 4 days later.
There was a recentish model, qwen next that was advertised as smth that can be run entirely on RAM.
They can ALL be run on RAM, theoretically. I bought 128GB so I can run GLM 4.5 with the experts offloaded to CPU, with a custom trellis/K quant mix; but this is a 'personal use' tinkerer setup basically no one but hobbyists will touch.
Qwen Next is good at that because its very low active parameter.
...But they aren't actually deployed that way. They're basically always deployed on cloud GPU boxes that serve dozens/hundreds of people at once, in parallel.
AFAIK the only major model actually developed for CPU inference is one of the esoteric Gemma releases, aimed at mobile. And the bitnet experiments, which aren't very big so far.
(In case it's not obvious, this is my special interest, and I'm happy to ramble on about how to set up 'niche gaming rig hybrid models' for anyone interested).
I suspect RAM may become increasingly useful with the shift from pure chat LLM to connected agents, MCP, and catching results and data for scaling things like public Internet search and services.
When I think of database system server software, a lot of performance gains are from keeping used data in RAM. With the expanding of LLM systems and it's concerns, backing data, connective ness, and need for optimisation, a shift to caching and keeping in RAM seems to suggest itself. It's already wasteful/big and operates on a lot of data, so it seems plausible that would not be a small cache.
Just in time for all those windows PCs that MS is trying to force people to upgrade
Who's bewildered? Of course this was going to happen. Everything enjoyable about life is being ruined. It's not surprising at all.
Wow I just checked the laptop kit I bought a month ago it is 50% more E: it's fucking over 100% more now wow it was a 64gb sodimm kit I got for 280AUD and now it is 700AUD fucking hell
Old computer blew up, had to buy a new one. Nice 128gb ddr5. Just mobo, mem, cpu. Cpu is a rhyzen 9 9700
The memory was well over 40% of the cost, wtf?
Laughs in ddr4....
It also got affected.
God fucking damnit, can't I even be poor in peace, JFC!
Indeed awful. AFAIU, it is stockpiling HBM memory ahead of next generation "GPUs", rather than existing product volume. I'm especially disgusted that prices in non-tariff countries are higher than in US. If this were to last, gddr5/6 computer ram would start to make sense. NVIDIA has been behind starving memory supplies for competing platforms (usually AMD) in the past.
This pricing is both huge inflation, but also huge drop in sales, because we have to wait for it to get sensible.
Far as RAM goes, it will become a good thing: it gives companies incentive to invest into the development of bigger RAM, more speed, and making the motherboard bandwidth big enough to handle it.
The next big generation of hardware will be much better IMO, simply because the companies will have to compete by their merits. The downside is not having enough supply right now, but once the logistics and tech is in place, even non-AI people will benefit.
No, it won't. The DRAM market is dominated by three companies, and they've colluded before. They get their wrist slapped by some government body, they promise not to do it again, and then they wait a few years and do it again.
The downside is not having enough supply right now, but once the logistics and tech is in place, even non-AI people will benefit.
Have you forgotten that they agreed to reduce production to stabilize prices? Capacity is not the real bottleneck.
Dang, now i know why micron stock has gone wild.
