As an old person, I recall a time period when there were lots of so-called RAM optimization software packages out there. It would be interesting to see if that sort of thing makes a comeback.
Yes, the operating system is far better at managing resources than back then, but I’d bet that lots of fun new malware will sprout up masked as such.
Expect an import ban on Chinese made ram to be coming soon.
The trump administration has determined that it is critical to national security that americans play all new games at medium graphics settings or lower.
Oh, I better then go ahead and buy them at high prices before it turns out in 2-3 years that he was wrong.
Get the Lenovo worker who said this. We shall dip him slowly into a volcano until he changes his mind.
Yeah, I’ll be there waiting for them to rotate their inventory at a loss when it does go down. Meanwhile, fuck Lenovo.
Why is never in quotes?
Because they are trying their best to keep it like this.
Oh course not. Why lower them when they can keep the prices high and pocket the profits? When you live in hell you can’t expect the devil to not profit on the vices.
The article ignores several points.
First, this is one of the conditions where capitalism actually works. Many players in the field dropped out because of the razor thin margins of the past. Fabs take years to ramp up, and are insanely expensive to set up, so getting in to take advantage of a temporary shortage was an unacceptable risk. Now there is a decade(s) long projected shortage, making the investment attractive again. Plenty of players have experience in Fabs, even though they are not up to date.
Also, the market is going to accept that slightly slower ram is quite fine in many applications, and they are easier to make. DDR5/6 is really not that important. I have an AM4 Ryzen 9 with 64Gb DDR4 that flies, I mean the thing cooks! This environment is going to make Chinese Fabs competitive in the mid-term, and give them the opportunity to catch up, especially since the Chinese government subsidizes whole sectors to catch up, and often surpass the west (see EVs, solar, airliners, etc.) maybe they’ll take years to get there, and maybe they won’t match the very top end, but they’ll take over.
We are going to have a shortage and obscene prices, but not as long or hard as doomsayers scream.
Another factor is that the AI bubble is going to pop. LLMs are a dead end, and are already at an extreme diminishing returns point. There is no way the major players are going to recoup investment, and the market will eventually wake up. Open source models are at single digit distance of the most powerful commercial models, so much of the resources are going to shift to in-house.
JEPA is one of the next steps in AI, and is way less hardware intensive. There are several new approaches to AI that are way less hardware intensive. LLMs are plain brute force approaches, and evolution makes efficiency a major goal.
Depends on government policy also I guess. What would most likely happen is for US gov to continue sanction Chinese companies like what they did to cxmt and ymtc to stifle competition and keep prices artificially high
This is your example of capitalism working? Then I don’t want to see it failing.
Yes. Capitalism is not bad, it’s actually a sane approach. The problem is when left unchecked, which is the state’s duty, but fucked up by politicians, in the pockets of oligarchs. Have you read Adam Smith? Both Smith’s and Marxs’ thesis fail beacuse they depend and assume inherent good in people, while in reality greed is the driving force in economics, and fuck all else.
AI will remain a massively parallel numerics affair with enormous data sets and monstrous memory bandwidth and network crossection. And accrding energy consumption. Jevon’s paradox will eat any efficiency improvements.
Only if LLMs are the only option. A paradigm change is coming. It’s like what happened when European and Japanese performance cars started to take on American muscle cars or SpaceX (yeah I hate the Nazitard too) started recovering rockets and reusing them, or PCs started replacing mainframe workstations…
Yeah any extra fabs able to produce ‘last gen’ DDR4 would have a big impact.
I have a work issued Lenovo Thinkpad P14S Gen6 AMD with a Ryzen 9 AI, 2TB nvme, and 64GB of GDDR5 RAM. It cost $2600 last October.
Went to buy more of them for other devs last week as their Dells are just hot garbage and are being refunded, who wants to guess what the price of the exact same machine though Lenovo directly again is now?
$6800.
Absolutely insane.
It’s just crazy.
Lenovo is not a manufacturer of ANYTHING below motherboards (pc or laptop) .
Their opinions need to be looked at through that lens and no other.
We have seen this story before. There will be a oversupply and then they will have to get rid of that inventory.
Lenovo apparently ignoring China then
Well, Lenovo is a chinese company.
But not a Chinese ram company.
Chinese no know China? Well shit.
Quite frankly, we abuse ram anyways. So much software uses way more ram than is actually necessary. I think this may be a catalyst to software fundamentals. Doing far more with far less.
It’s the only thing we are empowered to do, buy less ram and use software that runs smoothly with less ram.
Agreed, but still fuck data centers
Yeah, most people can probably be totally okay with 16GB (~150€ new, 50€ ddr4 used) too!
We seem to have forgotten that RAM has always been ludicrously expensive, except for a little while a couple of years ago.
16GB feels so low, but if we have to make due then the software we run needs to be crafted better. It’s entirely possible. Also ddr4 is plenty good. I wager no one can tell the difference between ddr5 and 4 or even 3 honestly
I was lucky I was forced to upgrade to 32gb right before the bubble, because my new job uses Jira with too many plugins.
That is a hillarious and deeply depressing reason for someone to have 32GB of RAM.
A work planner should work in 32M not 32G…
We should dismantle AI companies and their data center expansion plans and see if RAM still is so expensive.
Optimization will actually get much worse because software will be designed to run “in the cloud” on servers that have much more resources than the average budget pc or smartphone that 90% of users use for computing. You will own nothing etc etc
Software has been running in the cloud for nearly 2 decades at this point. But yes, I get your point. Their master plan is to get us all on terminals and use their clouds as a giant mainframe more or less.
Doesn’t have to be that way though. Not all tech needs to be the latest and greatest. So long as it’s secure and is feature complete who cares how it looks.
Ironically, with LLMs at our disposal, making new software is easier than ever. In the hands of skilled engineers tasks that took weeks take days. It’s more realistic now for a single dev to sit down with a goal like “let’s remake X software, but a 10mb memory limit”. We can prototype this kind of stuff faster than ever, so we can use the very tools causing this problem to solve this problem.
At work we were upgraded to 16GB last cycle. 8GB would’ve been plenty if not for all the shit that forcibly runs in the background. Now even the 16GB models are struggling, and I mean struggling sometimes. While my userspace apps use less than 2GB.
I’m going to guess you’re talking about Windows, correct? I work on a laptop (must be about 5 years old already) that I brought up to 64 GB RAM when I bought it (it was way cheaper than ordering with that, so I got it with 8 and then upgraded) only because sometimes I play on it. But when steam is not running and I’m just working, I’ve never seen it use over 2.7GB of RAM. Evidently, it has always run on some Linux distros (PopOS, Fedora, EndeavourOS and now CachyOS). The world is rigged to make all these things artificially expensive. Windows is a resouce-hogging malware that costs money, the computer parts manufacturers inflate prices to see if they can get away with it, the computer manufacturers go with that and then do the same, and we end up paying 10 dollars for what should otherwise be 1 dollar. As another person said, I also believe this will allow China RAM to catch up, which will end up flooding the market with same quality products, if not better, at way better prices. This will probably take a couple of years, but with their government subsidizing technology and most of their international markets the way they are in China, the funds will be readily available. Plus, China companies tend to enter difficult markets at a loss to take a good place if necessary, which is great for consumers. Just look at the blow they put on Mercedes, BMW and Porsche last year in Germany with BYD. Things will get better for us on the RAM front, it will just take a while.
Run less software then. You can do that today.
You don’t have to run less as long as you choose good software.
Also true.
But if you choose to run software that uses a lot of RAM, ask yourself why you haven’t created an alternative that doesn’t use a lot of RAM. If the answer is “I don’t have the time to”, then that’s probably also why the developer hasn’t made it use less RAM.
It’s more likely they are not incentivized to. When you are writing software for a living, typically there days the companies you work for prioritize delivery speed over everything else. If they prioritized memory constraints, software would use less memory.
When you are rewarded for features and delivery, you end up with shit like electron. Not to even begin talking about how a whole generation of developers learned to code for the web and never touch os level dev…
The answer to that question is that they’re either lazy, ignorant or both. :D
In all seriousness. I can imagine that a lot of developers who work on commercial products are given about 20% of the time and resources needed to make a good product. I don’t blame them for doing what it takes to not get fired.
That’s why libre source software is so important!
I think this may be a catalyst to software fundamentals.
It’s not that fundamental. It’s just corporations skipping the optimizing step and just shipping because that looks better for their project deadline and budget. As long as complaining is limited and sales don’t drop they don’t care.
Agreed. Only way this happens is either open source, or new companies that recognize it as a competitive advantage. Get new users to favor your software, they grow up to get jobs and advocate for it there. It’s a gamble and a long game though
This must be the wisest thing I’ve read in a long time.
They’re basically saying the AI companies are going to keep demand elevated to the point that supply will never catch up. It’s possible but with variables like public backlash, unrealistic power requirements, eventual financial and AI regulation, I would bet on a painful collapse.
no they are not. the article was clear. after the ai boom, the memory companies wont drop proces because you dont have a choice.
my new soap box - the government should regulate this industry because its very important to society and there should be prive caps.
Wrong line of thinking.
They want you to stop holding out and just buy their crazy high (along with everyone else) priced stuff so they move inventory and make some capital.
If their ram claim ends up true (I call bullshit) they make some sales and they go happily along.
If their claim ends up being false, they still made those sales and no one remembers or cares how “wrong” they were about it.
Saying this is basically a no lose guess for them.
Nah they’re saying the like 3 places that manufacture RAM won’t drop their prices after
You mean the companies that formed a cartel 20 years ago might do the same again? Couldn’t be.
The Chinese fabs should be producing lots of RAM by 2030.
There is literally zero incentive for companies to make ram and sell it cheap. The market is used to current prices, and by 2030 current prices will be looked at as cheap.
I doubt that, ASML won’t suddenly flood china with lithography machines.
Even if that’s true, I’m predicting either AI companies just buy all of that too. Or the US government doesn’t allow the import of it because it’s from China
Sucks to be american then
Isn’t it all from china already?
Old/current ram or new ram? Probably be moving on to DDR6 then
I’ll buy DDR4 if it means I can feed my family too
DDR5 will probably stick around for a long time if DDR6 is not affordable.
I’m still on DDR4 and it does what I need it to.
We could always just not, and use whatever RAM we can get. I’d rather have a thriving market with slightly worse RAM than motherboards that require a RAM no one can afford.
DDR2 Prices are up 60% as AI datacenters are slapping together whatever hardware they can get.
There is NO affordable RAM, and this is by design.
Nope it is for legacy systems which are not upgraded to new standard.
“Of course, today’s PCs don’t use DDR2, so we’re likely to see the impact of these price increases landing in areas like embedded systems, networking equipment, industrial controllers, automotive electronics, and other long-lived devices that were designed around it and are too costly to requalify on newer memory generations like DDR4 and five.”
I think these are false options. If there’s a thriving market for us there’s a thriving market for them
China will be running out of people soon. Their birth rate has been in collapse.
1.4 billion people will disappear in just 3 years?
Their birth rate is “In collapse” because they already have a gigantic population that is too big.
China will not collapse. It produces too many goods and its economy is too strong. Birth rates aren’t a problem to the point that they undermine manufacturing.
If someone tries to tell you the story that China will collapse and they seem credible, just reach out to me. I have a bridge for sale that will help you short the Chinese collapse.
The problem is they modify goods, not produce them.
Intellectual property shouldn’t exist. If China wants to “steal” ram designs and sell them for cheaper, then LeT tHe fReE mArKeT wOrK
That’s simply not how the current economy works.
That’s a wholly different topic.
But they’ll still have plenty of people by 2030.
And if they’re not as xenophobic as some other countries I could mention, they could easily solve the population issue through immigration.
And if they’re not as xenophobic as some other countries I could mention, they could easily solve the population issue through immigration.
Every Chinese book I’ve read screams at me nationalism and xenophobia, be it modern fantasy, 50 years old detective series or comedy.
And immigration to China is - currently - heavily restricted. Like damn heavily. And you will not become a permanent resident, and there is literally 0 chance of becoming a citizen, unless you’re of Han Chinese descent, preferably first generation, preferably looking Chinese, and you can’t have dual nationality (from last census in 2020 about 17.000 people ever got a citizenship, which is what. 0.0012% of population? And the permanent residency in 2017 was about 10k total).
My country, Poland (about 38m people), grants more than that every year.
Yeah, this will probably have interesting consequences down the line, but not in 2030, and maybe not 2100 even
maybe not 2100 even
Chinese government does not seem to agree with you, considering its:
- Pro natal policies
- Heavy research into automation being sold as “to fill in labour gaps” (even though the gig work is skyrocketing)
- Retirement age reforms increasing the retirement age
Chinese universities also do not agree with your flippant attitude and are literally alarmist about it:
Tsinghua projections (that accurately predicted peak of population) show halving the population (and that > 50% will be people over retirement age) by the year 2100: https://www.dess.tsinghua.edu.cn/en/info/1226/2472.htm
On the other hand the dude who was saying something about collapsing in 3 years also didn’t knew what he was talking about.
Yeah, but if they started right now I bet they could pump out some new adults inside of 20 years.
They literally can’t.
We’ll let the runaway inflation after the AI crash eat away at the currency’s value until RAM is back at the old value, even with the much higher prices, then wait for wages to catch up a bit. By then they’ll have a nice unsold supply.
People that think there’s going to be an AI crash are in for a rough future. The genie is out of its bottle and it’s never going back in.
You may as well be saying that there’s going to be an EV crash.
People aren’t saying AI is going to disappear, although that wouldn’t be a bad thing imo. When people talk about an AI crash they’re talking about the trillions in investments by the major companies that have been operating at a loss and have no monetization plan that could feasibly recoup even a fraction of that amount. Open source models are nearly as good as the big guys now, and nobody is going to pay hundreds or thousands a month just to keep using chatgpt or Gemini.
It’s not just the models that make the big players great, it’s the proprietary tools to use those models. Claude code, codex, seedance, etc are what everyone wants and what are so groundbreaking, not the LLM that they use.
EV crash, he he.
Nailed it
There’s always a collapse on the horizon, I believe many of these stories are worst-case scenarios or a way to help billionaires believe their own hype and swallow the turds to keep capitalism on life support.
I would bet on painful collapse, because the whole model is “winner takes all”, which means there is an awful lot of duplication. Even if it ends up more like a commodity with multiple players (because why pay for super powered AI for a task if there is a cheaper low powered alternatives?), the constant scale up makes no sense at all economically. We’re already well into diminishing returns with each scale up, and the models continue to be fundamentally flawed.
Lenovo are right that prices won’t go back to “normal” - I think there will be a huge crash in prices due to oversupply when the AI boom ends, and some of the big AI companies collapse.
We don’t even know if it’s currently being produced at greater numbers (as far as I’m aware). The only public info I’ve seen is that the data centers that haven’t been built will need all of the produced RAM based on handshake deals (not contracts). The RAM makers themselves are doubling down on the bubble by investing into the AI generators as well as increasing costs to insane levels in the process which surely reduces the amount of items sold.
Sounds like Lenovo is one of many companies pumping stocks.
I wouldn’t say “never,” but it’s very likely that RAM prices will not return to pre-AI (read: bullshit) levels. Many markets do this; hike up to crazy levels during a boom, come back down 80%, rinse and repeat.
The only thing that might put a stop to it is competition or the unicorn business that focuses upon everyday consumers and not purely profit (lol). I’m hopeful China is able to be a spoiler to this current tech hegemony, given general US hegemony is basically over, but the home computing market is probably fucked in the meantime.
Why wouldn’t China just charge the increased prices, same as the DRAM cartel? Primarily they want to make money.
Unless they offer below-cost RAM for a few years to drive the competition out of business before jacking up their prices of course.
Because that’s not how you make money from the household consumer market. We’re already there, and people aren’t able to buy RAM.
You don’t make money from product that doesn’t sell.
I’m hopeful China is able to be a spoiler to this current tech hegemony
As awful as that sounds, I bet that’s what will happen.
That’s what I’m guessing. The consumer market is a multi-billion dollar industry. If Micron, Samsung, and SK Hynex are too big for it now, that’s them leaving money on the table for others. Smaller companies will kick up, like those in China, and they’ll gladly take the money left for them. It’ll just take a while for them to get there.
The thing is, are they really leaving money on the table? Higher prices means less customers, yes, but also higher profits per unit. We don’t actually know right now how many people would be willing to pay these new prices: if it costs twice as much but half the customers are still fine buying it, then profit stayed the same. And with no choice, I suspect much more than half are going to pay double.
Of course, the people who wouldn’t do that are a potential market, but they are also a much lower profit margin. Comparably, that’s probably not that much for these people.
There’s a whole segment who are going unserved because of those high prices. Gamers across the board and enthusiasts, the entire consumer market, because RAM is too expensive. Someone else figures out how to make RAM at scale and undercut the fake inflation of the cartel could make billions easily. That’s money sitting on the table.
maybe i’m naive, but if it’s so profitable to make RAM in the long-term, why wouldn’t competition emerge? I get not investing in the startup costs just for a bubble, but that’s not what we’re talking about here.
Capital/time intensive start up costs make it a barrier to entry. This is why the prices are so high. Supply is inelastic because the producers know this is a bubble. If they do the capital intensive thing and the bubble pops before realizing the additional capacity, they are left holding the bag.
Yes, that’s literally what I said about bubbles. The assertion in OP is that RAM pricing won’t go back to pre-bubble prices. If that is true, RAM manufacturing will be incredibly profitable post-AI-bubble and competition should emerge eventually.
It takes double digit billions to start a manufacturing plant and that’s when you already have people who know what to do.
Most countries can’t really afford this in their budgets and I’m saying countries because it’d be a stupid endeavour for most private enterprises to even attempt. CXMT (DRAM) and YMTC (NAND) absolutely are sponsored by China, which is the only real way to get one of those companies going these days.
Google, Apple, etc could start their own memory companies if they wanted to. But it’s a hell of an expense to justify to your investors.
Making RAM isn’t like making a shirt or a suitcase v it’s an extremely specialised and extremely expensive business, and this should be apparent given there are literally only 3 manufacturers in the world.
Making ram is incredibly hard and expensive, that’s why there are only 3 companies in the entire world that do it. For a new company to attempt to do it now, they’d have to outlay tens/hundreds of billions of dollars and a decade before they see a cent of revenue.
And ASML making the lithography machines for basically everyone… You can’t just “make more” ram without scaling ASML too.
