RAM? What RAM? The biggest makers are reportedly now sold out all through 2027

Memory chipmakers Samsung, Micron, and SK Hynix have reportedly sold out their DRAM inventory through 2027, driven by AI companies consuming nearly 70% of DRAM production capacity, according to Taiwanese website Digitimes. SK Group's president said AI demand for RAM is expected to rise 60-100%, with storage demand up 50-50%, worsening supply shortages for PC and mobile memory. The three companies were also sued for price fixing in the U.S. District Court for the Northern District of California on June 25.

I hope you don't like RAM or graphics cards or anything like that. Prices are expected to rise again, and stock is expected to vanish through 2027. That's according to a report from Taiwanese website Digitimes https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=1&id=0000763847 DVY7YHX65GMLYZ6UQEEMP , which a rough translation notes that inventory has already been sold out all the way through 2027 from the likes of Samsung, Micron, SK Hynix and others so we expect this sticky situation to continue well into 2028 at this rate. All thanks to the rise of AI companies which have been sucking up "nearly 70% of DRAM production capacity" and so "DRAM supply of PC and mobile phone applications will be relatively limited". From their report they noted that the SK Group President mentioned that RAM demand from AI companies is still expected to increase by "60~100%" along with storage demands continuing to rise at "50-50%" so your drives aren't safe from this either. We already had Micron pull out of consumer memory https://www.gamingonlinux.com/2025/12/the-ram-price-and-availability-situation-is-going-to-worsen-as-micron-pull-their-crucial-consumer-business/ , and make some rather long-term deals https://www.gamingonlinux.com/2026/06/expect-ram-prices-to-stay-high-with-micron-locking-in-deals-for-5-years/ that will continue to strain supply chains. So the situation really does just continue to get worse and worse . In other news, in case you missed it, the likes of: Samsung, SK Hynix and Micron have been accused https://www.law360.com/articles/2493985/major-chipmakers-sued-for-price-fixing-amid-rampocalypse- of fixing memory supply and prices. This lawsuit was filed in the United States District Court for the Northern District of California on June 25th. It's going to take quite a long time for supply to catch up with the demand, as building new fabs doesn't exactly come cheap and takes a good amount of time to get going. 🌐 External Sources: digitimes.com.tw https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=1&id=0000763847 DVY7YHX65GMLYZ6UQEEMP , law360.com https://www.law360.com/articles/2493985/major-chipmakers-sued-for-price-fixing-amid-rampocalypse- All posts need to follow our rules https://www.gamingonlinux.com/index.php?module=rules . Please hit the Report Flag icon on any post that breaks the rules or contains illegal / harmful content. Readers can also email us https://www.gamingonlinux.com/email-us/ for any issues or concerns. External Link: You need to be for an in-depth look at how much of a shitshow this whole thing is. logged in https://www.gamingonlinux.com/index.php?module=login to view this.Here's a quote from the beginning of the piece: As I wrote in the Hater’s Guide To The Memory Crisis, the sheer scale of Microsoft, Google, Meta and Amazon’s spend on AI data centers has led to a massive supply chain crisis and price-gouging from the triopoly of Micron, SK Hynix and Samsung, with Micron alone bumping prices for DRAM by 60% in its last quarter, shooting up the price of every single kind of RAM possible, at a rate increased by the amount of GPUs and servers that hyperscalers buy. To give you a sense of how memory hungry AI is, single 72-GPU GB300 NVL72 AI server has over 20 terabytes of high-bandwidth memory used almost exclusively in GPUs and other AI chips , and 17 terabytes of the LPDDR5X RAM used in mobile devices, and a gigawatt data center has thousands of those NVL72 servers or something similar . Moreover, that high-bandwidth memory that AI GPUs use requires more wafer space during manufacturing — further reducing the amount of manufacturing capacity for other kinds of memory. This naturally creates a vicious cycle. The more AI servers that hyperscalers buy, the more demand they create for RAM and high-bandwidth memory, which increases the price of RAM and HBM, which makes the AI servers more expensive, which means hyperscalers need more money, and because AI has yet to provide meaningful improvements in revenue or cashflow, they’re forced to raise more debt. See more... See more... See more... Quoting: GuestBtw, "AI" Crawlers and "AI" Agents cause now for the first time in history more internet traffic than all humans together.Bots, before whole "AI" BS, already gobbling 50% of traffic in 2010s. Select originators/target countries had over 60%. See more... Quoting: GuestJust want to point out, everyone who is using so called "AI" regularly that includes the usage of Google and default DuckDuckGo is part of this problem. While it is true that these companies are creating this problem in first place, every regular user is justifying the behavior by using these "tools".You can blame people who use cloud gaming too. I have 56GB of RAM and 48GB of VRAM with Geforce NOW Ultimate. How long do I have to wait until I can afford to buy a house from selling it? Therefore: please join the AI hype so prices keep rising BTW: want to know why AI will never be what you desire? Have a look at this original text is in german, but at least at translating AIs are "OK" ..... most of the time : Gödel & Turing - Limits of AI External Link: You need to be logged in https://www.gamingonlinux.com/index.php?module=login to view this. See more... Quoting: GuestWhat I'm hoping for, and this is already proving to be true, is that we are EXPANDING supply. New factories in the US for example from TSCM and Samsung. AI continues to expand, and ram supply continues to grow which will lower prices.Zero impact, I don't think they even do DRAM or NAND, but micro-processors instead. The whole problem is that you cannot start new lines that fast, it can take 5 years. Hyperion for those that wonder what I am on about 😇 See more... Quoting: GuestThe AI bubble will deflate just like the .com bubble in the early 2000s and supply will increase. The sucky part is that it will take a bit of time to do.Too bad the planet will be out of water by then: a typical data center uses more than 1 million liters of water daily, and add to that the water used by the power plants... See more... Quoting: GuestThat isn't true at all. Demand for AI is growing, and doesn't look like it'll be stopping anytime soon.It will stop as soon as people have to start paying the true cost of the service. For now we are still in the phase were the drug dealer gives out doses for free, things will get interesting when he starts demanding payment. Quoting: GuestThe AI bubble will deflate just like the .com bubble in the early 2000s Balloons can deflate, bubbles usually burst... like the .com bubble of 2000. See more... Quoting: GuestI'm already paying for the service. 200/month.First I hope you use it for work, I cannot imagine someone with an average salary paying 200/month dollars/euros? for personal use of AI. Second I hope your salary has gone up by more than 200 since you started using it. 🥰 See more... and because AI has yet to provide meaningful improvements in revenue or cashflow, they’re forced to raise more debt.This is the elephant in the room. RAM chips may be bought throughout 2027 but they sure AF aren't paid for. It's IOUs from top to bottom. The AI industry is like a junkie borrowing money from its friends. Sooner or later there won't be anyone left to borrow money from. See more... See more... Quoting: Dmitri Seletskihot damn, i am so lucky i upgraded to 64 gig of ddram with 7900xtx. i am hunkering down, covid style, for next a few years.I did a full system upgrade last year, so I think I'm good for a bit. See more... Quoting: GuestJust want to point out, everyone who is using so called "AI" regularly that includes the usage of Google and default DuckDuckGo is part of this problem. While it is true that these companies are creating this problem in first place, every regular user is justifying the behavior by using these "tools".I am surprised that a Linux forum community might expect common folk to make good technology decisions after witnessing 20+ years of Windows hegemony. We can use the past to inform the future - AI will dominate. It will be commonplace. Hardly anyone except for people like you and I will give it a lick of thought. Quoting: GuestI'm already paying for the service. 200/month. See more... Quoting: GustyGhostWe can use the past to inform the future - AIWe have to convince people to lick more.willdominate. It will be commonplace. Hardly anyone except for people like you and I will give it a lick of thought. See more... Quoting: GuestNo, you are mistaken, Demand for AI is growing, but it's growing at a much slower pace than the supply is. Think about it, how many AI providers does the average user need? Two, three, maybe four? Only within the LLM space there is Google, OpenAI, Meta, Anthropic, DeepSeek, Qwen, Mistral, xAI, Cohere, not to name countless other LLMs being developed by various countries with support from governments thinking that owning AI is a strategic goal. Compare this to other services and you'll see that the supply definitely far exceeds the demand. For example, how many search engines does the world actually need? Google alone covers more than 80% of the market, then you have Bing and that's almost it; the rest are all just small players.Quoting: Savor592Oh boy. I sure can't wait for the news of all these new AI start-up golddiggers failing horribly, as AI-supply heavily outweighs demand.That isn't true at all. Demand for AI is growing, and doesn't look like it'll be stopping anytime soon. What I'm hoping for, and this is already proving to be true, is that we are EXPANDING supply. New factories in the US for example from TSCM and Samsung. AI continues to expand, and ram supply continues to grow which will lower prices. This is almost a rule: whenever there is a new technological breakthrough there is guaranteed to be a bubble, i.e. supply far exceeding demand. This is because when there is a new technological revolution many investors will think it's the right time to invest, but none of them will be able to know for sure how many other investors are also thinking the same way, so what do you do? Do you wait and miss out, or do you take the risk and invest? No one will want to miss out on a golden opportunity, so it's guaranteed that the supply will outweigh demand. It happened with the internet bubble despite the demand continuing to rise and it's happening now, and investors know by now that we are in a bubble but it's too late to back out because they've already invested tens of billions, so they have no way other than try to delay the bubble burst through false information or whatever until they either secure a place among the few or find a way out before the burst. See more... Quoting: Guestand they won't be able to ask for the true cost of the service if you have enough RAM to run the models locally, so they have to buy out all RAM in the market to make sure you don't have the option of running models locally.Quoting: GuestThe AI bubble will deflate just like the .com bubble in the early 2000s Balloons can deflate, bubbles usually burst... like the .com bubble of 2000. Quoting: stormtuxIt will stop as soon as people have to start paying the true cost of the service.Quoting: GuestThat isn't true at all. Demand for AI is growing, and doesn't look like it'll be stopping anytime soon.It will stop as soon as people have to start paying the true cost of the service. For now we are still in the phase were the drug dealer gives out doses for free, things will get interesting when he starts demanding payment. Quoting: GuestThe AI bubble will deflate just like the .com bubble in the early 2000s Balloons can deflate, bubbles usually burst... like the .com bubble of 2000. See more... far less "compute" per query. But those big US companies are the ones building most of the data centres, or getting some other companies to build them on the proposition that they will be using them once built. Meanwhile, the "AI" services offered by Google etc. that show up "AI" answers every time you do a search . . . have no revenue at all. Arguably they indirectly create revenue by stopping search users from leaving the Google search page to go to places on the internet that actually created the answer in the first place, and thus mean Google can charge more for ad space right on the search results page. But that in turn means that ad space on actual web pages will be worth less . In the long run, obviously if you are the dominant player in online advertising, and you do something to cannibalize the general prosperity of the online ecosystem, that online ecosystem will in the end be paying less for ads overall. But even in the short run, I can't see that Google can be making money on this stuff after expenses. Now Google, and Meta, and Microsoft, have a lot of revenue, massive profits, outside of "AI" stuff. They can carry a big loss on "AI" for a long time. But I expect when OpenAI and Anthropic go under and "X" drags SpaceX into the mud and all the hype goes away, the likes of Google and Microsoft will chalk it up like Meta did with the Metaverse. So again that will be that much less "compute" and that much less need for data centres. So, sure, when the bubble bursts there will still be some Chinese companies using 1% as much compute. But the huge capital expenditure will dry up, the infinite appetite for memory and graphics cards will disappear, and the majority of use cases, which upon examination have generally turned out to be useless cases, will go away. LLM AI will become a much smaller, more niche thing. Oh, and there will be a major recession and hardware will get cheap again. See more... we cant purchase new once what we have break... See more...