I’m even more confused by the economics of the frontier model businesses. Even if they got preferential prices, they are still paying an inordinate amount for their infrastructure.
If it wasn’t economically feasible without VC/Nvidia money 3 years ago, how is it possibly economically feasible now at 10x prices for things like memory?
And watch … $5 says they have nothing to do with the ram because even if there was enough power there’s no use case for this much spend
If I had to guess they’re just buying up ram to keep others from having it same way meta, Google, etc hired up engineers to keep others from having them
I’ve been wondering this if you buy up all the memory you can ensure the competition cannot compete with huge investment raising the bar for competition. Of course this all sounds all conspiracy theory. Though without a place to power this or put it makes me wonder this.
It also would be a huge win for the memory companies to agree to this, considering they will make more money with less supply. Seems interesting to me.
Also imagine people having their own cheap 2TB rigs at home for their own usage/inference, or a multitude of small providers serving/sharing capacity thanks to cheap available hardware.
But we cant have that, as the big guys are hoarding everything for themselves, so that you have to pay them instead.
I don’t think there were some grand conspiracy to buy up the world’s ram to keep other people out, it is probably more likely that they just really thought they would need it.
However, I do feel like this is the type of conspiracy that unfiltered Claude would come up with.
It's pulling the ladder up behind them. OpenAI bought up more co tracts for wafers than they could even use. But now that means everyone else has to fight over what's left. This makes it too expensive to compete with OpenAI. No upstart can really afford to build infrastructure.
The situation is absolutely awful for anybody who needs significant memory or storage (or devices incorporating such).
Even people who think they have enough for the next few years already could end up with an unpleasant surprise if they e.g. have a stick or two of RAM go bad or a GPU fry. I have a few machines with aging sticks that I'm hoping will manage to hold out until when/if things improve.
I've seen some fail. Might have had to do with the fact that the computers were literally flooded (river came up faster than the sandbaggers could keep up) while running.
I've seen computers brought back to life after flooding by lying in rice for a week, but they weren't turned on when the water hit (and not all survived).
I've had a few RAM sticks fail, but the ECC usually caught it first. Also had one case that somehow the stick slightly unseated itself; dunno how but it was out of service for months until someone opened it up and reseated the stick.
...granted though I've also seen a backpane on a server fail. I didn't think that was even possible until that point.
I recently had part of a 32 gig DIMM go bad on a 128 gig system. One of the larger VMs was spontaneously rebooting for no apparent reason. I was able to find the addresses with MemTest86 and map out 512 megs of the affected DIMM with the Linux kernel memmap option.
Don't send it back. Since they can't easily replace it with new hardware, they will probably refund you the price of purchase. Which is normally a very good deal, but obviously not in these circumstances. I've seen it happen with GPUs as well.
I had a 2012 mac book pro that burned out 2 different crucial memory sticks. Would have bricked the machine if it was the soldered-on type, but it was the last model apple sold that have swappable memory. Crucial had a lifetime warranty so they replaced them for free.
I used to game on it and I think the design just couldn't handle the cpu and dgpu being active for long periods, it would get extremely hot in the area close to where the memory was.
on older PC bios would perform a memory test - you may remember those - and that's where it would mark a bunch of pages dead/unusuable and it was not uncommon at all.
Granted, I've been building my own computer for 30 years and I've only seen it happen ONCE.
But I've seen plenty of other failures:
- Two hard drives -- One got dropped on the floor, so it was no surprise, the other was showing degraded performance and SMART showed some scary numbers and I was able to replace it before I lost data.
- One GPU, replaced under warranty.
- One AIO water cooler, replaced under warranty. Somehow, the water vanished from the loop. I'm guessing a microscopic leak that evaporated as fast as it leaked.
- Two power supplies -- One randomly exploded. Just started making popping noises and shooting sparks out the back. At autopsy, I determined that the fan failed (it was hard to spin manually) and it overheated. The other was just being overloaded. I had just gotten a new GPU and the PSU wasn't big enough for it.
- Several case fans.
Never seen a CPU or motherboard failure, even when my AIO water cooler was failing and my CPU was constantly at thermal limits. Never had an SSD/NVMe failure.
Anecdata: a coworker found a single bit flip corruption in a file he wrote in his computer. He ran memtest86 on his memory, and which found one of the two sticks was defective. The computer otherwise ran perfectly fine for years; and we would have never noticed if the file format didn't have built-in checksums.
How do you know they weren't faulty? Blip flips are usually silent unless you have ECC memory or are actively testing for faulty cells.
Somewhat recently, Firefox implemented memory testing in their crash reporter and found 10% of reports were due to bit-flips[0].
If it weren't for reading that HN thread, I would never have being prompted to check and find the faulty memory in my laptop. It would have gone undetected and continued to corrupt whatever was stored in the affected cells. I did run memtest86 not long after I bought the laptop, so sometime between running that first memtest86 and the 12 months after, it developed the fault.
Non-ECC memory sometimes fails and there's a good chance you'll never know.
> I've been using computers for more than 40 years now and the one component I have never seen fail is RAM.
I think over the decades I got one RAM disk fail. And another one didn't fail but, although I bought a pack of four sticks (2x x2), was detected (by memtest, when building the rig), as having another model name than the three others: even the shop who sold me the RAM was confused (don't know how that happened).
The usual component that I've seen fail the most would be the PSU? (before I started buying quality ones). One NVMe drive (an "ADATA") died on me even though it was nearly new.
PC assembled with quality parts usually last a very long time.
Oh well, it's fairly common.Maybe not in a singular/several rarely used (and not monitored) desktop computers, but.... it's common once you have quite a few under your care.
I've seen every single component of a system fail, the least I've seen fail has been the CPU itself (I've only seen one fail out of thousands).
Memory seems to fail pretty commonly overall, less common than harddisks but more common than motherboards.
The issue is that I'm used to using ECC ram, which fails loud when it's actually bad.. consumer memory won't tell you unless you can't boot.. and everyone disables the startup memory testing too.. (in fact, I think it's disabled by default for the last 10 years because people want to boot quickly).
I was so sure that a CPU wouldn't suddenly fail that I ended up replacing every other part in my computer before getting a new CPU and realizing that was the issue.
I also have been using PC's for a long time, thirty years. The only time I have seen ram fail during use was within the last 5 years. My 10th gen platform and 32 gigs of ram DDR4 and one of the sticks failed, this was before the ramapocalypse of course, the sticks were replaced under warranty. But I have had systems in my shop with failed ram. it happens, but its rare, unless its cheap ram, less rare.
I have had a RAM stick start frequently failing ECC checks after a couple years use. I also had the VRAM on a GPU fail, but you could blame that one on overheating from the nearby compute core. I agree is rarer though.
I'm saving a ton of money by simply not upgrading. I used to do upgrades every 1-2 years just to keep up on things, pass the old box down to my son or otherwise donate it. Now I intend to keep my current setup for 7+ years. Its 2 years old next month. I may even stay with it longer. I just see no reason to upgrade with things how they are.
I considered a Mac Studio a couple months ago. Price was up but might have pulled the trigger if not for the delivery date that was almost 4 months out. Sale lost. I suspect a lot of people are like me right now and we'll start seeing significant drops in consumer spending on electronics the next 2-4 years minimum.
I have a 2021 M1 pro and see no reason to upgrade it... ever?
I do web & general java work on it and it's plenty fast even using a memory hog like IntelliJ (which is typically disk bound anyway). Claude tells me that a new macbook would feel about 1.8x faster in my CPU-bound cases, which is not common.
If memory prices returned to normal and if local inference gets over the hump, I could see upgrading for local AI agents, but I'm fine waiting for a good long while on that.
I say all this as someone who, five years ago, was a perpetual upgrader: immediately bought new product and got excited for next product. Wonder if the future will see the end of that culture in tech and, looking back, it was a little silly and wasteful anyway.
I too have an M1 Pro and it does everything I need to do. I bought it in 2023, so it's 'only' 3.5 years old now.
Maybe this is the tipping point where hardware companies can't rely on the upgrade cycle for sales any more. I've always upgraded at least every three years, but with the combination of high RAM prices and good enough performance from an older machine, there is no need to.
I'm in the exact same position. I want to upgrade, but when I interrogate the reason, it's all just empty consumerism. This computer does everything I need.
There isn't much to upgrade to if all you want is a desktop for gaming, a non-workstation work box or just a decent laptop. Much like smartphones, there just hasn't been much innovation of late. A midrange gaming desktop from mid 2023, something like rtx 4070 and ryzen 7800x3d, can still play pretty much any video game on max settings and high framerate at 1440p, and has sufficient performance for most workstation-esque tasks as long as you have enough ram. Similarly, the only thing my spare macbook m2 lacks compared to m5 is hardware av1 decoding - the performance is worse, but I don't really notice it in normal usage, and macOS 27 runs perfectly well on it.
Local AI will crush any machine, of course, and if you need a 512gb ram box you'll probably need to sell both kidneys, but you're well past the average consumer territory with those.
So, fingers crossed nothing breaks down and that some of the funny money goes into R&D which will eventually trickle down to consumer devices in a few years (16TB gum stick SSDs at sane prices pretty please) instead of endless datacenter GPUs and SSDs that will be tossed into a shredder in x years.
I'm running a 12 year old i5 3.2ghz CPU with 16 GB of memory, an SSD and an AMD RX 480 (8 GB) GPU. Currently using a 100% scaled 4k monitor and a 2nd 1440p monitor. I keep the machine on 24 / 7. It's great for software development along with video + image editing and casual gaming. Everything feels fast and snappy.
I will use this until the hardware stops. There's a 0% chance I'd buy hardware at the current prices.
I have a baseline iMac Pro and custom built tower with similar specs (3.2Ghz Xeon W-2140B, 32GB RAM, Vega Pro 56 and i7-6700K, 32GB, 5700XT respectively) and they're both perfectly responsive. For many day to day things the difference between them and newer machines is negligible. For other things, my 5950X tower and M5 Max MBP run circles around them.
Both the 5950X tower and M5 Max MBP are deep into diminishing returns territory regardless, however. Aside from some extreme niche workloads like LLMs I can't imagine needing anything more powerful. They both have a ridiculous amount of headroom for the overwhelming majority of anything most people might want a computer for.
Yep, I have an M4 MBP company issued laptop for work and in my day to day there's not a enough of a performance win to feel like I'm missing out for most tasks.
Sure, the Slack web app might initially load half a second faster since there's a lot of CPU bound JS executing but generally speaking I never feel like "wow this sucks" on my main machine. It's kind of the opposite, for a number of things my desktop machine feels faster than the MBP for overall "using the computer", with development heavy tasks (running Docker based web apps, browser tabs, various tools open, streaming music, etc.). It was like that when I ran Windows 10 Pro with WSL 2 and now native Linux with Arch + niri.
Even CPU bound tasks on my 12 year old box that I do a lot, such as editing a blog post on a static site with Hugo takes ~100ms to get live reloading with over 1,000 posts and drafts. I'm sure if I were constantly doing huge CPU bound tasks I'd have issues but I don't have that use case.
After recording 500+ videos, encoding isn't a problem too. If I record a 30 min video, it takes ~20 minutes to render but it's so easy to fit that into other areas of my life, like I'll record + edit a video and then render it while I walk.
My desktop PC turns 20 next year. I've never seen any need to upgrade it, went out of my way to build it with silent parts and it rarely if ever goes outside a single-digit CPU load (before anyone asks, software development and a bit of CAD work). I'll upgrade it when something in it physically fails, which means I'll need to replace it with new hardware with 5-10x the specs just to get the same performance.
My laptop is currently about two years into a typically 5-year lifespan, limited by when the keys start to fail rather than any need to upgrade it just because. The RAM in it is now worth more than the rest of the laptop.
I think it's probably for the best, the hyperconsumerism around electronics was getting out of hand. Computers are fast and have been fast for a while now. An iPhone 12 is probably more than fine to do iPhone things, if Apple didn't sabotage their old phones with poorly-written poor-performance software.
See this is what I don't get. This effects big organizations like this, there should be a class action lawsuit for these criminals LLM and Memory companies.
it is difficult to motivate hyperconsumers with the narrow band of improvements that can be made with new LLM models and also are specific to smartphone OS integration and not superseded years in advance by datacentres
I have a MacBook M1 and I recently had to downgrade from Tahoe to Sequoia. Why? First of all they force you to download 7 GB of AI training when you’re not even using their AI. And secondly on Tahoe just watching a video with nothing else running off-line my MacBook was asking for 7 W of power. On Sequoia that drops down to 2 W maybe three.
It’s sad to think they think of themselves as an environmental company, I don’t know if they do, but this update to Sequoia literally doubled the amount of M1 max that are running on the planet in terms of power use.
Did more research and apparently it’s a bug Apple never fixed that causes the deletion to only work sometimes. To fix it you need to run this Terminal command in recovery mode:
> Computers are fast and have been fast for a while now.
Absolutely. The only reason I want to upgrade to an iPhone 18 next year is because the 16 doesn't have MIE (Memory Integrity Enforcement) and it feels ridiculous to upgrade so quickly to the 17, which does have MIE. Same thing with a Neo 2. But mostly I buy surplus gear. I've saved a small fortune this way and 8% compounding returns (VT and chill) will help me retire early.
Personally I have zero reasons to upgrade from IPhone 13 until the software stops working. I hope it will last me several years more. Beautiful hardware.
Can you provide evidence for your claim? Most people take this statement at face value and I don’t understand why. I do some volunteer work supporting elderly people that have issues adapting to a digital only world and I am always pleasantly surprised by the performance of their old iPhones all the way down to the 8 (old phones they get from their kids)
Every time I upgrade iOS past a few major versions there is significant slow down. I don’t think it’s done on purpose, it’s just that software grows more demanding over time.
Nah. Adding more features to software doesn’t make software run slower. The problem is that software developers only optimise enough to make it run fast on our own computers. Because we keep getting new computers, we get lazier over time and software gets slower.
Electron is the most obvious example of this sort of thing, but examples are everywhere once you start looking.
Yes. Microsoft is scrambling at the moment because the macbook neo only has 8gb of ram and runs great. Windows 11 can't compete on that hardware because it's so bloated.
I find it strange given microsoft made windows 3.11, which ran fine on computers with just 4mb of ram. Windows 11 is of course more complex than windows 3.11. But not so much that it should need literally 2000 times as much ram.
In some ways, the memory apocalypse is a great thing. The windows start menu should never have been made in javascript and react. I'm hoping our industry uses this as a push to make our software small and fast again. If ram is 5x more expensive, we should aim to make 2gb of ram be more than enough for regular computing tasks.
I wonder if the computing requirements of windows scale exponentially? It looks like it.
I suspect if we graphed them, they’d pretty perfectly match the computers that were available. Ie, I think windows is optimised just enough to run on the low end hardware of each generation.
I agree, it’s not a natural phenomenon per-se and it is getting worse. I think this consumer hardware problem will actually help get us back on track, to a degree.
Apple admitted to slowing down iPhones to keep the batteries from failing, and agreed to stop doing it and created a new battery replacement plan. I’ve personally experienced the benefit of getting the battery on Apple devices replaced which made them behave like new again. There’s a reason old Apple devices hold their value a lot better than competing manufacturers.
It was a pretty terrible controversy at the time but I think the battery health settings pane we got out of it was a good compromise and a net positive. What would be even better would be fully user-replaceable batteries.
To be more precise, the phones weren’t slowed down to keep the batteries from failing, they were slowed down to maintain stability and stop random crashes. As batteries age they lose their peak voltage capability, so you’re basically undervolting your components unless you slow them down.
They hold value now. They didn’t if we go back to before those lawsuits. Apple can blow smoke wherever it wants, but that 500 million loss isn’t about battery health.
As an aside, I have this lovely tower from France, barely used, scrap price should be very high. Priced to sell?
It's wild seeing so many comments about updating every 1-2 years.
When I build a PC, I pay a little extra because I want it to last a long time. I aim for 5 years, and in reality it usually ends up 7 years (life happens, need the money, etc).
I am now in year 8 of my current PC. I only had money to replace it this year, and am now stuck with it due to the RAM prices. It's going to really suck for me if my RAM fails in the next few years.
Exactly. I probably spend 1.5x the money on a really solid build, and it's paid off - I'm still running just fine on a 10 year old desktop (overclocked i7-6850K). The only component I've upgraded (other than hard drives) is the graphics card (to a 2080Ti), and it's still running all the games I play just fine. I was planning to upgrade it on the 10 year mark, but now I'm just going to make it last a few years longer until RAM pricing becomes less silly again.
I run Qwen 3.8 at just under 2 tokens/second on a Mac Pro 2013. Why? Because I walked 8 miles to school in the snow, uphill both directions, every day.
I've always done 5-7 year cycles on machine I've owned, but this time around its getting quite complicated as my last refresh was 4-5 years ago. So I'm kind of stuck wishing my current hardware both meets the requirements and just survives another half a decade, or my next upgrade, even if only partial, will be VERY expensive...
Similar situation here, my desktop is from 2019, with only the GPU upgraded since then. On the other side, I'm also much less interested in modern games these days so perhaps I'll do fine with a 10+ year cycle this time.
I would absolutely love if some startup came up right now to sell RAM at normal prices to everyone else and put Crucial et al out of business for price-gouging everyone just because they can.
DRAM fabs are multi-billion dollar investments which take years to come online. That's the whole reason we're in this situation, there is no quick fix available.
China is making central decisions for its RAM makers (and who can complain now that the US isn't a legitimate free trade defender). I think their strategy could go in any direction to either get a small portion of still gouging or a large portion of a market with somewhat stale specs.
DDR4, and from some comments I have seen around here of only marginally acceptable reliability. It may still take a few years if until they have all the production processes fully under control.
Earlier this year I finally bit the bullet and grabbed myself a 5080 for my gaming PC. I wanted to push things a bit more than my 3060ti allowed, and I figured it probably won't get any cheaper any time soon.
The PS5/Xbox Series console generation is supposed to be wrapping up soon even though it barely feels like it got started. I don't sense there's appetite for more powerful consoles anytime soon (I don't believe the PS5 Pro made waves like the PS4 Pro did), and jumps in console performance are what determine when games start targeting higher system requirements. So I like to think I can get a good number of years of use out of this.
I built my machine in 2022 with 64GB of memory and upgraded to a 4090 at MSRP in 2023 so at this point I'm good but I fear the back end of the bathtub curve. I currently expect to have to hold on until 2029 and at that point the mobo and CPU will be 7 years old, and the GPU almost 6.
same-ish. In 2023 128 gigs of RAM and a 4090 and 4x 4TB ssds. I just looked up prices and 4090 alone is like 2/3 of what I paid for the whole system. Crazy times.
This is one of the rare situations where an out-of-control homelab & PC building hobby is actually useful. With over 250GB of RAM & 90TB of storage I will be fine for a while, especially as long as DDR4 era machines remain relevant.
My issue is my digital hoarding. I've always said it I "collect" things digitally vs. physical, so I can scratch that itch without causing a disaster for my life. A half rack of JBODs is a lot cleaner and overall less expensive to operate than deciding I need to collect every single Lego Technic set ever produced or something.
The problem is I also operated under the prior "storage always gets cheaper over time" and was right at about 80% full when all this insanity hit. I'm now at 90% on my largest ZFS array (~280TB) of spinners... and I get to decide if I want to drop a few paychecks on the next vdev upgrade, or start deleting things.
Not to mention I've been holding off replacing a bunch of old 1TB SATA drives (19 years old now!!) as my scratch partition, as I was going to wait until 8TB SSD got really cheap and just buy two of them for a RAID1 for the use. And now I need those drive slots to upgrade the mass storage pool.
Talk about backing oneself into a corner at the wrong time!
I'm at 97% of my 160TB and I need to start making hard choices :-/ I have Unraid, and an old 4TB drive sitting in a drawer which I could plug into the array. It's old, and I planned to throw it out, but fuck paying these prices for a 26TB HDD right now.
If you bought new it should be warrantied. Most sticks are essentially lifetime support, but you’ll get a refurb. I definitely wouldn’t buy secondhand now, unless it was a very good deal. Corsair won’t honor transfers unless you’re somewhere that legally forces them to.
And that's the really fun part of all this, that weak substitutes will also become more expensive. But on the bright side, I've got dozens of sticks of DDR3 RAM that I'm hoarding like Pokémon cards. Any day now, we'll get desperate enough to start using them again!
Been long past that time, probably 15-20 years past. It's not even the languages its the 500 different layers of abstraction and abominations like Electron.
Deep down it's all just people telling stories. If the AI sector's story is setting the price, and yours is not, then theirs is the real real one, not yours.
It's a totally broken way to practice economics, but alas, it's ours until we change it.
It's very unlikely to have any meaningful effect. CXMT DRAM will also be hoovered up the moment it hits the market at the same inflated prices. Even if they had the capacity (they don't) why flood the market when you can just print money?
They are building fabs on track to get online towards the end of 2027, full capacity production expected for 2028. These are serious pieces of very complex technology that take years, not months, to build.
I bought an RX 6600 for $450 during the GPU shortage a few years ago. I know the AI boom is different than the crypto/COVID boom, but still. Don't be me.
Even with AI pressure I'm willing to bet there's no need for prices to be this high. There has to be some price gouging/fixing going on. Same thing for nand and GPUs. Or else they're doing it to try to pressure people not to use local compute instead of the "cloud"
Memory manufacturers have gone through several cycles of lawsuits for market collusion at this point. It's gone from an open secret to just outright flaunting in our face.
Until the penalties involve actual jail time or a fine 10x larger than the profit they made from doing it, they couldn't care less.
I noticed that resellers on ebay of used server equipment that I keep an eye on in Germany pulled all their DDR4 from surplus and used server racks and now list it separately for sale, at jacked up prices. Where you used to be able to see a used Gigabyte server listed with "128GB" "256GB" etc listed in the title, they now all say "0GB". They took extra effort, to pull used ram from older servers, and list it separate, just out of the greed of potentially capitalizing on this market.
The problem is... now for anyone looking to pickup a slightly older used/surplus server rack - you have the go and buy the (now marked up) RAM separate.
So... did they list the "0GB" racks with removed ram cheaper you ask?
Nope. Many of them seem to be basically the same price as they were or even higher than they were, except now you're paying for RAM separate.
I bought a surplus Gigabyte server about 2 years ago off ebay for a home project, and I just went and the same seller still has some for sale, even older now than they were when I grabbed one, but they are 2x the price, and they don't come with any RAM, whereas I got one already populated as they were from the factory. Once you add that back in literally looking at 150-200% markup in the last 2 years on what was just surplus computer equipment that is 5+ years old.
> So... did they list the "0GB" racks with removed ram cheaper you ask?
Could this be an indication that there's a supply shortage/price increase of the remaining components in the rack? Or do you interpret it as people just trying their luck out of sheer greed?
Greed? Resellers aren't charities, good for them if the value of their stock goes up. Besides, people buying server equipment aren't exactly starving, I bet most of them can either simply not buy the server, or cut some other expense.
Why would I voluntarily choose to sell the hardware I have for less than the going price? The odds of it getting picked up by a scalper and resold are extremely high. If I was sitting on a goldmine of old server hardware, I would absolutely be ripping out the RAM and selling it per stick in this market.
This is why the companies should be aiming for a soft landing to this fad (if they know that they can't make the ROI that makes the industry worthwhile).
same. still rockin a 2016 era dell xps 13. I was about to pull the trigger on a few laptops but I always psych myself out because I assume prices will drop and I can pick them up secondhand. big mistake...
500% in a year is just insane. I knew RAM prices were going up, but $3k+ for 128GB of DDR5 makes me wonder how much of this is actually AI demand, and how much is this just manufacturers taking advantage of it. They have done this in the past after all
To take advantage they'd have to do some form of price fixing which in most places is a very big footgun.
I honestly don't think they need to fix their prices, and it's possible that selling to consumers is a secondary market for them compared to datacenters, like Nvidia selling GPUs to gamers. We know that's not where the money is.
Seeing some companies actually exit the consumer market is evidence for this.
> To take advantage they'd have to do some form of price fixing which in most places is a very big footgun.
When parent commenter says "They have done this in the past after all" that's exactly what they're talking about. Micron, Samsung, Hyinx (prior to SK Group acquisition), and others all participated in RAM price fixing in the early 2000's.
I don't understand the distinction you are making. If you have a massive excess of demand ready to pay any price, and a limited supply, of course the sellers will take advantage, that's how any market works. They can do that because of the lack of supply.
Over the long term demand will rationalise and supply will adjust. But in between there is a squeeze, and the right price is the maximum price buyers are willing to bid for the limited supply. Sucks for other buyers, on the other hand it ensures demand destruction, i.e. that the limited supply goes to the buyer who need it the most, other buyers will delay their purchase and do with what they have.
In an absolutely free market? sure. But we don't live in an absolutely free market, the memory manufacturers have refused to budge on creating new fabs until recently out of caution and to be able to create a demand excess.
They have slow rolled rollouts and everything, if there wasn't CXMT they would probably still wouldn't have budged on building new fabs for a while longer.
You can absolutely price fix/manufacture a shortage.
Ever heard of merchants setting fire to crops after hoarding to sell at a higher price?
The issue is the distributer and producer are the very same people this essentially means producers have no incentive to rationalize supply, they can make much more money of the long drawn out shortage.
I think everyone is aware now that you don't build a fab by snapping your fingers, and you need to amortise it over many years. The said manufacturers are just out of a massive slump in demand during covid, when they were selling at a loss, and these cycles happened many times in this industry. It is understandable that they don't jump all-in on the AI bandwagon, which, god knows where it is going and for how long.
If you want evidence that no one predicted this, look at WD. They disinvested from sandisk up to early this year. Now sandisk is worth more than WD. Doesn't seem consistent with a concerted strategy to gauge prices.
> You can absolutely price fix/manufacture a shortage. Ever heard of merchants setting fire to crops after hoarding to sell at a higher price?
Fine. Show me which manufacturer is cutting production to corner the market. I don't see that.
Big manufactures have absolutely cut production of lower margin products and it's not like they can re-purpose all of the equipment used for it for HBM yet. edit They can definitely focus more resource because of it on HBM though.
Things have improved a bit since govts are starting to get angry.
Tldr; no one is cutting the production of high margin stuff, but low margin stuff can go and die, because it's not a free market it's a stock market these companies want to show off better margins in every unit there is no reason to keep inefficient things running when not doing it pays more.
They just don't want 30-40% margins anymore. That is not free market that is quite literally serving their own interests because they know there is a monopoly.
All big memory manufacturers are heavily supported by the govt, that's what has made them the monopolies they are Micron supported by USA (they crushed Japan's memory industry for it), and Korean giants are more tangled up with the govt than well most people in US would realize.
Somehow people think free market means anything goes, which honestly is where we stand today so I don't blame anyone.
But using govts to crush competitions and build yourself up on subsidies and similar isn't really free markets. China hasn't been able to grow a significant memory chip industry yet because of lack of EUV machines and similar technologies blocked by the US, which is also not "free market".
As for WD, and Sandisk separating I feel like this was more mutual and they new HDDs or SSDs being together was a distraction for both. Which feels more strategic, think of it like this, the shareholders who had the shares how hold shares in both and together they might have likely been worth less so shareholders made money.
Aka good for free market. I don't see how that relates to the actual memory crunch issue more broadly.
Margins pre-AI were more like 5-15%. Even if the margin was 5%, if prices increased 5 times but product cost didn't move, then margins are now 80%. That's not a 30-40% product, your argument doesn't make any sense.
> But we don't live in an absolutely free market, the memory manufacturers have refused to budge on creating new fabs until recently out of caution
...that's how the market works. This is called capital discipline and is standard across industries. If the increase in demand is temporary, they would be spending 5-10 years of profit on facilities which will only lose money. The same works in non-capitalist economies too, you don't want to waste resources on stranded assets and many still believe this RAM demand is temporary and the house of cards will collapse.
> the memory manufacturers have refused to budge on creating new fabs
That's EXACTLY how a free market is supposed to work. The existing memory manufacturers are FREE to not build new factories, and anyone else is FREE to build those factories in their place.
> If you have a massive excess of demand ready to pay any price, and a limited supply, of course the sellers will take advantage, that's how any market work
The question is how quickly the supply adjusts. In a free market, the suppliers would respond to the price signal by quickly expanding their production to meet the elevated demand.
But a market captured by a cartel will be much slower to expand the supply.
So, the test isn't "how much is price going up" it's "how aggressively are the RAM producers expanding supply production"
Memory fab build out times are years not months. The existing ones already run 24/7. If these companies could pull more supply out of thin air they would, but they cannot.
Cartels aren't known for massively increasing production to satisfy all demand. Far from it I'd say, they're more likely to expand supply to such a level that they can still reap massive profits. Very few countries/companies have the knowhow and willingness to be able to build more.
OPEC is a perfect example of this behaviour. They set hearly production goals to keep prices high. Or the Quebec Maple Syrup Producers organization that limits the amount of syrup a farmer can produce and sell in a year to keep prices high.
But are the memory chip producers acting like a cartel? In the past they have not. When there was a shortage, they increased production as rapidly as they could build the new fabs, crashing the price of memory.
Wasn't that bad for business last time? I could see them erring on slower build up because of how it went before, if they're not sure how long this demand will last especially
Which is exactly what's happening. There used to be a lot more then 3 makers but up till very recently memory was a comedically low margin business you didn't go into.
Well guess what :) They've done it before, they will do it again.
"To date, five manufacturers have pleaded guilty to their involvement in an international price-fixing conspiracy between July 1, 1998, and June 15, 2002, including Hynix, Infineon, Micron Technology, Samsung, and Elpida.[1]"
That was 24 years ago. Half the people commenting on this board aren't even that old. How many of the shareholders on the boards or executives at the helms of those companies do you think are still there?
As others have noted they've been found colluding previously, more than once. Even admitted guilt and some people went to jail over it.
And the reason the market is so small is because of players who did as you propose, the DRAM market crashed on them (not due to them but due to broader market dynamics), and they went out of business.
So there is, quite perversely, strong incentive to collude if possible and if not to abide by an unstated agreement to be extremely slow to expand DRAM capacity.
That said...
SK Hynix has two fabs under construction, one set to open later this year. One for 80k wpm, another 360k wpm. A third was recently approved to start construction.
Micro has an overhaul of Fab 6 in Virgina near done and slated to produce first wafers this year. It also has a large expansion in Boise nearly completion for 300k wpm. They're also building a "megafab" site in New York (my guess is >1m wpm) and expanding at their Japanese plants (40k wpm).
Samsung has two under construction, P4 for 100k wpm.
Nanya - historically a secondary player focused only on producing non-leading-edge DRAM is making a play for the leading-edge market with a new fab dedicated to DDR5 and HBM.
It appears they've all decided that the demand is sufficiently sustained that building new fabs is worth doing... but it takes time. If my guess is right Micron is making the largest play with their New York site at 1-1.5m wafers per month. That would give them a large chunk of the market.
OPEC's goal is as much profit as possible. They're not thinking in the short term but long term. If as much possible means over the span of multiple centuries then that's what they're going for.
I think we're in agreement here. I don't think memory manufacturers are interested in a market disruption even though profits are incredibly high for a few years. Make it too profitable for too long and others will arise but keep the supply at a steady pace with decent margins and they will remain the only ones on the block because of massive upfront cost.
China is, to my knowledge, the only country that has so far had the willingness to seriously invest. Not necessarily for profits but for independence reasons. I don't see many (or any really) other countries following suit.
Maple trees are living things and have a right to reproduce, you can't fault a bird for shitting maple seeds all over my farm planting maple trees everywhere.
I saw a documentary(forgot the name, sorry) about that that said, no you can't choose not to join. I forgot if it was about this organization but it was about maple production in Canada where a farmer wanted to do their own thing and they were intimidated and all sorts of nasty things happen to them because they didn't think the org was just and wanted to sell freely. Very interesting documentary and an eye opener that even farmer unions can act like organized crime when there's money involved.
> you can't fault a bird for shitting maple seeds all over my farm planting maple trees everywhere.
If only building a ram FAB was that easy :).
The best hope is that cxmt gets going quicker than expected and is able to satisfy a decent chunk of the Chinese demand for RAM which would - in theory - reduce the demand for ram from the few other producers.
It looks like you're forced to participate by law once you sell over a certain volume.
> Backed by the Canadian civil courts, the federation has the monopoly for selling Quebecois maple syrup on the wholesale market, and for exporting it outside the province. Producers are only allowed to sell independently cans of less than five litres or 5kg to visitors to their farm.
https://www.bbc.com/news/business-35028380
You're correct in theory. I'm just not sure we have a cartel situation. Micron is a US company, and it controls 25% of the global market. I see no evidence that they nor SK hynix nor Samsung are intentionally limiting production.
I do see people arguing, "they had years to increase production - why didn't they?" They've been remarkably candid about this: because RAM production has been losing them money in the past. Because they don't have a cartel, prices were on the floor. Investing in more production would have been corporate suicide.
Even now, there is considerable risk that the bubble will burst, and all this investment in new production will be flushed down the toilet. RAM production is a brutal business.
I'm curious what the limiting factor is for "years not months" when building these fabs. DRAM is a commodity, there are only threeish major manufacturers, all have huge amounts of capital, multiple nations bend over backwards for new semiconductor production projects, and there seems to be unlimited money going into purchasing these things.
Even when it comes to manufacturing the tooling/testing infrastructure, it's just a question of money aiui.
Large barriers to entry are anathema to competitive markets, so the conclusion here is still that firms are benefitting from the lack of a competitive market.
Memory fabs take years to build, they cost a lot and there is no guarantee that the demand will stay as high years in the future. It's all very complex.
There are several Chinese fabs that they could contract out production to. In fact that's exactly what Apple tried to do, before the U.S. government went full protectionist and added the two memory manufacturers to the banned imports list. There's also the Texas Instruments fab that's underutilized right now. The extra capacity exists, it's just Micro and SK Hynix are throwing their money around to stifle anything that could jeopardize their spike in profits that are historic not only for their industry, but for manufacturing as a whole.
Even if the memory market was completely competitive the time it takes to expand supply is measured in years. There was no way for them to serve the AI demand without reducing supply given to everyone else in the short run
> There was no way for them to serve the AI demand without reducing supply given to everyone else in the short run
I didn't say that there was a way to do it in the short run. I laid out what the test is to determine if there's cartel-like behavior or competitive-market-like behavior.
I will say, in a market with only 3 players and very high cap-ex cost, my prior is that you're likely to see more cartel-like behavior than competitive-market-like behavior, but that's still to be determined.
> But a market captured by a cartel will be much slower to expand the supply.
Oh yea, governments can fix by putting 100 billion in memory fabs right now, and when demand crater in decade or so, let tax payer eat the cost of those fabs. At least it will benefit poor computer users like me.
Generally speaking, if demand stays high, and existing suppliers don't increase supply, the high prices create an incentive and opportunity for new market entrants to increase the supply.
Yes, "quickly" is obviously industry-dependent. If they were fabricating something extremely easy to manufacture, "quickly" might mean months-to-a-year. Fabricating something very complicated like memory "quickly" might mean years.
That said, even though we might have to wait years for new supply to hit there market, we probably don't have to wait years to see signals of that. We will likely know that manufacturers are starting to build new supply long before we see the chips hit the market.
In the current environment unless people start decommissioning datacenters en masse supply will be locked up till the early 2030s IMO.
Even if you started a new fab today, it's 5 years till it comes online. The great hope here is that China's DDR5 yields represent a high reject rate of an enormous amount of new capacity that they'll dump into the market to get a sovereign capability in that area.
As has already been pointed out further down, OPEC is the perfect example of this. Things don't really work that way when the time scales are so extended. What we need is fracking, but silicon.
This system is too complex, the few individual players too big, with too many artificial distortions for the adjustment to equalize out in a way that's anything but years-long (it may have shock-waves that extend past the decade), and highly destructive to everyone who isn't OPEC, or their largest customers, in this analogy.
The brunt of it will be paid by the people further down the food chain.
Where are you hallucinating that this is not a free market? There is no cartel, there is a massive massive investment and skill up cost to building a large scale fab.
When you don't understand how anything works, everything looks like a conspiracy.
> According to the one-count charge filed in San Francisco's federal court on Thursday, Park conspired with unnamed employees from other memory makers to fix the price of DRAM sold to computer makers from April 1, 2001, to June 15, 2002. The government says the move directly affected sales to U.S. computer makers Dell, Hewlett-Packard, Compaq, IBM, Apple Computer, and Gateway.
> I don't understand the distinction you are making.
It's pretty simple. In a correctly regulated market the sellers are in competition with each other so there is a limit to how much they can increase prices, since they'll just lose business to their competitor.
In a poorly regulated market that allows competitors to collude with each other to raise prices by the same amount, they'll inflate prices much further.
I'm sure there will be no large scale leverage campaign against a country who is enabling profiteering at this level. No defense agreements removed, troops removed from the country, or large tariff increases around failed investments.
I'm sure it will only suck for the buyers. This has no effect on inflation and probably will be overlooked by the current admin.
>I don't understand the distinction you are making. If you have a massive excess of demand ready to pay any price, and a limited supply, of course the sellers will take advantage, that's how any market works. They can do that because of the lack of supply.
The question is how much of this demand is actually based on AI, or just an excuse to fleece regular buyers (and AI is not really the actual demand driving such prices).
Memory manufacturers have driven up prices via collusion many times in the past (and paid fines for it), without AI or some similar high demand vertical market existing.
But I don't think it's the case now, see my answer elsewhere here for why.
In a market with healthy competition (no collusion), this generally can't happen. Every single vendor is incentivized to lower their prices slightly at all times, with a floor at cost plus margin.
Floor is below cost if you think you can put one of your competitors out of business before you run out of money, and you'll make the money back after doing so. And this includes even after fines.
>In a market with healthy competition (no collusion), this generally can't happen.
We're talking about the memory market here.
The "healthy competition (no collusion). Every single vendor is incentivized to lower their prices slightly at all times, with a floor at cost plus margin" is the fairy-tale version they teach at 101s.
>that the limited supply goes to the buyer who need it the most
Everything else you said is accurate, but this is not. Supply will go to those most able to pay, not those who need it the most. It is not a moral allocation, which your language implies.
There have been multiple cases of collusion and market manipulation in the RAM industry throughout history, the most notorious or documented from the late 90s. This is not a misunderstanding of econ 101.
FWIW I have a friend who makes hardware and it's extremely difficult for him to secure memory at any price. Because AI people have locked up so much of the upcoming supply it's often not a question of how much it will cost but of finding someone who will sell you memory at all. If this were price gouging I would expect he'd have no trouble finding allocations at exorbitant prices but that doesn't seem to be the case.
What is also interesting is no one predicted this in 2022-2025 as LLMs were gaining widespread use. It just suddenly happened--"boom!" memory shortage in mid 2025 to early 2026,and it caught everyone by surprise. It was understood that the biggest bottleneck was "compute", not the memory. Even experts, such as popular AI bloggers, industry leaders, and top podcasters failed to anticipate this. No one on twitter either. I spend a lot of time online and I don't recall, afik, anyone before 2026 saying to buy memory stocks.
Sam Altman laying claims on 40% of the world’s RAM wafer production is what took everyone by surprise. The supply shortage likely would have ramped up more slowly otherwise.
More importantly that did take subterfuge: there's no way a contract that large would've been agreed to by both manufacturers if they knew the other was signing it.
The actual deals were hours apart and secret for that reason.
It's hilarious watching the conspiracy theorists do their usual thing when there was an actual conspiracy, it was perpetuated by the inner circle of a US company and announced and publicized in full once it had achieved its goal.
I think Leopold of "situational awareness" fame predicted this, tried to make it widely known on podcasts and x, and still started billions worth of hedge funding based on this.
> I think Leopold of "situational awareness" fame predicted this, tried to make it widely known on podcasts and x, and still started billions worth of hedge funding based on this.
Ironically enough it's because of its bet on RAM companies that all of Situational Awareness' publicly trading stocks got bought by the highest bidder (Citadel) when the (leveraged) shit hit the proverbial fan and Situational Awareness created a hole of tens of billions.
Situational Awareness was not forced to sell its share in Anthropic for, by contract, investors in Situational Awareness have no right to force the sale (before this month of September I think) of shares in companies not publicly trading yet.
Citadel bought those RAM stocks right at the end of July, during the dip. Amazing timing for for example MU bounced back from $746 back to above $1000 (now around $950).
Now Situational Awareness may still do exceptionally well: they've got $5 billion or so left in Anthropic AIUI.
Eventually 2028 and 2029 will come around and at some point today's data center builds will be (economically) full.
In addition to that there are extra fabs build and (for example) China can use the high margin to catch up with their processes without breaking the bank.
The average hatching egg costs a few dollars. Can go as high as nearly $20 if it's something remarkably rare and useful in terms of breed/quality.
An egg takes 21 days to hatch.
Then it's 18-24 weeks to start laying their own eggs. That's ~$5-15 in feed over the period, maybe 40-50 pounds. There's utilities, labor, land, etc., but your incremental cost to add a hen if you have the space is minimal. My numbers won't reflect a lot of operators, as scale matters.
Out of a typical 1.5-3 year lifespan that had a startup cost of a few dollars per, paying back its return by about month 6-7, that's not too bad. It's only worth doing financially in this economy at ludicrous, arguably inherently inhumane scales, but that's what it is on paper.
Not quite the same path to course-correction as a silicon shortage.
>and how much is this just manufacturers taking advantage of it
"Manufacturers taking advantage of it" wouldn't take it to 500% without AI, because it's way beyond the price elasticity points where they make up lost sales at higher prices.
Those prices lose most consumer sales, and are only viable if there's constant high demand that's relatively insensitive to price (so, not regular folks, but e.g. AI companies).
Taking advantage of Trump admin, AI hype, and setting themselves up as "too big to fail" for bailouts when demand eventually collapses
See the tariff thing where Bessent bought rights to refunds
It's all a really low tech con perpetuated by weak elders who need blue pills to simulate virility. Odd how they complain about the kids virtual reality.
Oh no it's not either; it's intentional petty bullying
I almost upgraded my computer a year or two ago, but decided to hold off for a bit longer.
I've always been fine with being "behind" on the latest and greatest. I don't mind watching a movie, playing a game, buying a cell phone, etc, that was released a few years prior. If it was really all that great when it came out it should still be good later on.
Buying slow usually saves a.good bit of money when prices go down over time... but this time they went way up and now I'm kicking myself on it. Not upgrading the computer for now, but it is a bit overdue since I had already put it off for many years.
Except staying on the edge costs exactly the same as not, when you take the resell value of the components on the second hand market into account. The problem is that some can afford to lock their capital in the hardware (via one-time investment) and some do not.
R9700 will lose less than 30% of value in the upcoming 3 years. Give or take. Yet people buy ancient cards for way more than 30% of R9700's cost.
> ...when you take the resell value of the components on the second hand market into account.
When taking the resale value of used kit on the second-hand market into account, one must also take into account the hassle of selling on the second-hand market.
For me? Personally? That hassle absolutely eliminates any money I would get from any plausible sale price. If I'm getting rid of used kit, it's going to be as donations to people I personally know who will come and pick it up at the agreed upon time.
Yes and I seriously doubt gps insinuation that resale could make up for the price premium of top of the line components when they were consistently beat by the next years significantly cheaper mid range components.
I upgraded a year ago and I bought a bunch of extra RAM because it was "so cheap", why not. Man did I get lucky, I got it right before the price hikes started.
Same, my desktop is a top end full of ram beast that I thought I overpaid for because motherboards had gone up. Bought everything before the surge. Past me was damn lucky. Wish I had bought some stock tho.
I ordered a new CPU on sale a while back, right when the memory prices started spiking. I didn't even know RAM was going up in price. Realized the CPU needed a new motherboard and ram so I ordered the motherboard, started looking at ram and realized it would cost almost as much as the Mobo and CPU put together. Cancelled the orders because I wasn't intending a big expensive upgrade.
What I have is still fine, though I do regret cancelling now.
I thought about upgrading my macbook, but I have the M1 Max and it still doesn't feel slow, so I decided it'll probably last another two years. That said, now I feel like this price raise is maybe permanent. If so it will definitely eat into computer makers profits, people will not be upgrading nearly as often.
last year i mis-bought a ssd on amazon (wrong size or model i can't recall) so I asked for a return, surprisingly amazon wanted to save on transport that they offered to pay me to keep it (i don't want to spread stupid claim but maybe the full item price), i was confused and worried it would create issues down the road so i sent it back thinking i'd buy the proper item the month after.. and now i'm running duplicate search programs to save space on drives.
If you want even more bad news: display panel makers also want to raise prices
"Due to the continued rise in the prices of core components globally, the production cost of monitors has increased significantly, and we can no longer absorb this pressure on our own."
It would strange if just when a lot of articles mention computing / social networks are detrimental, governments start passing laws to forbid platforms and/or phones for kids, and now price gouging flares up everywhere, this would entice people to just stop using these devices and maybe enjoy natural taper-off.
People do not want to talk to their devices because they are in the room with other people and they dont want to disturb them. I think their device is going to be a flop.
It's going to flop because everyone is already able to talk to Google Assistant to manage things, and outside of very specific circumstances (e.g. driving) everyone is still tapping screens instead
Ram, hdd, ssd, even blu-ray writers are expensive. I purchased mdisc blu-ray writer for $70 in 2021, I see people are trying to sell the same device for $500+ wtf
Is it a Pioneer model? They stopped making their blu-ray drives which are at least perceived to be better quality than the competition, so NOS prices have gone way up.
I have written a llm compression quantization library called glq because of the high ram prices. Glq uses qtip Trellis quantization which are efficient at low bpw 2-4 bits. As a gamer and ai developer I want to squeeze more out of the same hardware. Glq runs with vLLM.
After RAM prices climbed 200% between August and November of last year, a store associate at Micro Center encouraged me to buy some RAM sticks if I needed them because "prices might continue to rise for a few years". I told him that was ridiculous and nobody would tolerate it if prices got higher. How wrong I was...
If it makes you feel better. The US government looks like it might start removing some of the 28k troops that are stationed in South Korea over this. No reason to keep them in a country that enables this sort of behavior.
An admin upset about inflation driven negative economic sentiment and a country that is slow rolling a couple hundred billion in investments. Current events.
What else do you know about the inner workings of the current administration… and seriously reflect on why your answer isn’t “lol what? Absolutely nothing!”
I was facing a similar situation when I upgraded my system about a year and a half ago. I wanted to upgrade from 32GB RAM to 64GB. The kit (DDR5 6000) was $179. I knew that I might want to add an additional 64GB later on, so I had to decide if I should wait for prices to drop further.
I decided to get the extra RAM. Boy am I glad now. If I thought this was just a temporary spike, I would be tempted to sell the extra memory (I don't really need it yet), but I don't see prices falling anytime soon.
I built a new main rig in November because I saw the memory prices going up. I overpaid about AUD$110 above its RRP. The same kit is now AUD$610 above RRP from the same retailer.
Here's hoping I can wait out the current shortage before any upgrades (I typically only upgrade every 5-7 years or so anyway) or replacements, because I do not plan to pay the new price.
421 comments
[ 0.22 ms ] story [ 32.7 ms ] threadIf it wasn’t economically feasible without VC/Nvidia money 3 years ago, how is it possibly economically feasible now at 10x prices for things like memory?
https://www.tomshardware.com/pc-components/dram/openais-star...
If I had to guess they’re just buying up ram to keep others from having it same way meta, Google, etc hired up engineers to keep others from having them
According to them, demand for tokens already exceeds supply. Coding agents are very popular but burn through millions of tokens per hour.
The reasoning models that have powered breakthroughs in AI for math, coding, etc require substantially more compute than base LLMs.
But we cant have that, as the big guys are hoarding everything for themselves, so that you have to pay them instead.
However, I do feel like this is the type of conspiracy that unfiltered Claude would come up with.
Even people who think they have enough for the next few years already could end up with an unpleasant surprise if they e.g. have a stick or two of RAM go bad or a GPU fry. I have a few machines with aging sticks that I'm hoping will manage to hold out until when/if things improve.
All the other bits in a computer I've seen fail.
...granted though I've also seen a backpane on a server fail. I didn't think that was even possible until that point.
I used to game on it and I think the design just couldn't handle the cpu and dgpu being active for long periods, it would get extremely hot in the area close to where the memory was.
Granted, I've been building my own computer for 30 years and I've only seen it happen ONCE.
But I've seen plenty of other failures:
- Two hard drives -- One got dropped on the floor, so it was no surprise, the other was showing degraded performance and SMART showed some scary numbers and I was able to replace it before I lost data.
- One GPU, replaced under warranty.
- One AIO water cooler, replaced under warranty. Somehow, the water vanished from the loop. I'm guessing a microscopic leak that evaporated as fast as it leaked.
- Two power supplies -- One randomly exploded. Just started making popping noises and shooting sparks out the back. At autopsy, I determined that the fan failed (it was hard to spin manually) and it overheated. The other was just being overloaded. I had just gotten a new GPU and the PSU wasn't big enough for it.
- Several case fans.
Never seen a CPU or motherboard failure, even when my AIO water cooler was failing and my CPU was constantly at thermal limits. Never had an SSD/NVMe failure.
After the second, HP replaced all that we had from that batch.
HDDs fail every 3-6 months, there's a lot of them (3-4PB).
One PSU.
One CPU/motherboard. (I let the HP engineer handle that one.)
One NVMe drive.
Somewhat recently, Firefox implemented memory testing in their crash reporter and found 10% of reports were due to bit-flips[0].
If it weren't for reading that HN thread, I would never have being prompted to check and find the faulty memory in my laptop. It would have gone undetected and continued to corrupt whatever was stored in the affected cells. I did run memtest86 not long after I bought the laptop, so sometime between running that first memtest86 and the 12 months after, it developed the fault.
Non-ECC memory sometimes fails and there's a good chance you'll never know.
[0]: https://news.ycombinator.com/item?id=47252971
I think over the decades I got one RAM disk fail. And another one didn't fail but, although I bought a pack of four sticks (2x x2), was detected (by memtest, when building the rig), as having another model name than the three others: even the shop who sold me the RAM was confused (don't know how that happened).
The usual component that I've seen fail the most would be the PSU? (before I started buying quality ones). One NVMe drive (an "ADATA") died on me even though it was nearly new.
PC assembled with quality parts usually last a very long time.
Common scenario:
200 PCs at least 6 years old, often over a decade old, come in on pallets.
We pick up the first 50, stack them on benches, and power them on.
At least 10 wont post.
Of that 10, reseating memory and other little jobs fixes 5.
Of the remaining 5, replacing or removing memory gets 4/5 them to post.
Testing those sticks in other PCs confirms the RAM is dead.
Into the scrap it goes.
The remaining PC is dead mobo, PSU or needs a new CMOS battery.
Memory seems to fail pretty commonly overall, less common than harddisks but more common than motherboards.
The issue is that I'm used to using ECC ram, which fails loud when it's actually bad.. consumer memory won't tell you unless you can't boot.. and everyone disables the startup memory testing too.. (in fact, I think it's disabled by default for the last 10 years because people want to boot quickly).
The only thing i've never seen fail is the Case lol. Air cooler too if you don't count the fan.
I considered a Mac Studio a couple months ago. Price was up but might have pulled the trigger if not for the delivery date that was almost 4 months out. Sale lost. I suspect a lot of people are like me right now and we'll start seeing significant drops in consumer spending on electronics the next 2-4 years minimum.
I do web & general java work on it and it's plenty fast even using a memory hog like IntelliJ (which is typically disk bound anyway). Claude tells me that a new macbook would feel about 1.8x faster in my CPU-bound cases, which is not common.
If memory prices returned to normal and if local inference gets over the hump, I could see upgrading for local AI agents, but I'm fine waiting for a good long while on that.
I say all this as someone who, five years ago, was a perpetual upgrader: immediately bought new product and got excited for next product. Wonder if the future will see the end of that culture in tech and, looking back, it was a little silly and wasteful anyway.
I'm also using an iPhone mini 12 and it's fine.
Maybe this is the tipping point where hardware companies can't rely on the upgrade cycle for sales any more. I've always upgraded at least every three years, but with the combination of high RAM prices and good enough performance from an older machine, there is no need to.
Local AI will crush any machine, of course, and if you need a 512gb ram box you'll probably need to sell both kidneys, but you're well past the average consumer territory with those.
So, fingers crossed nothing breaks down and that some of the funny money goes into R&D which will eventually trickle down to consumer devices in a few years (16TB gum stick SSDs at sane prices pretty please) instead of endless datacenter GPUs and SSDs that will be tossed into a shredder in x years.
I will use this until the hardware stops. There's a 0% chance I'd buy hardware at the current prices.
I have a baseline iMac Pro and custom built tower with similar specs (3.2Ghz Xeon W-2140B, 32GB RAM, Vega Pro 56 and i7-6700K, 32GB, 5700XT respectively) and they're both perfectly responsive. For many day to day things the difference between them and newer machines is negligible. For other things, my 5950X tower and M5 Max MBP run circles around them.
Both the 5950X tower and M5 Max MBP are deep into diminishing returns territory regardless, however. Aside from some extreme niche workloads like LLMs I can't imagine needing anything more powerful. They both have a ridiculous amount of headroom for the overwhelming majority of anything most people might want a computer for.
Sure, the Slack web app might initially load half a second faster since there's a lot of CPU bound JS executing but generally speaking I never feel like "wow this sucks" on my main machine. It's kind of the opposite, for a number of things my desktop machine feels faster than the MBP for overall "using the computer", with development heavy tasks (running Docker based web apps, browser tabs, various tools open, streaming music, etc.). It was like that when I ran Windows 10 Pro with WSL 2 and now native Linux with Arch + niri.
Even CPU bound tasks on my 12 year old box that I do a lot, such as editing a blog post on a static site with Hugo takes ~100ms to get live reloading with over 1,000 posts and drafts. I'm sure if I were constantly doing huge CPU bound tasks I'd have issues but I don't have that use case.
After recording 500+ videos, encoding isn't a problem too. If I record a 30 min video, it takes ~20 minutes to render but it's so easy to fit that into other areas of my life, like I'll record + edit a video and then render it while I walk.
It runs Factorio and Zelda BOTW, why do I need more? :)
My laptop is currently about two years into a typically 5-year lifespan, limited by when the keys start to fail rather than any need to upgrade it just because. The RAM in it is now worth more than the rest of the laptop.
It’s sad to think they think of themselves as an environmental company, I don’t know if they do, but this update to Sequoia literally doubled the amount of M1 max that are running on the planet in terms of power use.
sudo rm -rf /System/Library/AssetsV2/com_apple_MobileAsset_UAF_FM_GenerativeModels
How I love modern software lol
It’s telling the software cycle is a scam, when new major software versions slow perfectly fine machines with zero value added features.
Absolutely. The only reason I want to upgrade to an iPhone 18 next year is because the 16 doesn't have MIE (Memory Integrity Enforcement) and it feels ridiculous to upgrade so quickly to the 17, which does have MIE. Same thing with a Neo 2. But mostly I buy surplus gear. I've saved a small fortune this way and 8% compounding returns (VT and chill) will help me retire early.
Also the coming iOS 27 release supposedly has some big performance improvements for all supported models that people have verified in the beta releases even with the iPhone 11: https://www.macworld.com/article/3160623/apple-finally-figur...
https://www.bbc.com/news/technology-67911517
Nah. Adding more features to software doesn’t make software run slower. The problem is that software developers only optimise enough to make it run fast on our own computers. Because we keep getting new computers, we get lazier over time and software gets slower.
Electron is the most obvious example of this sort of thing, but examples are everywhere once you start looking.
Apple also seems hellbent on fully utilizing all available space on the low end model. IOS+ System Data is about 30GB.
The original iPhone had a 4GB model and had space left over for the user to use!
Yes. Microsoft is scrambling at the moment because the macbook neo only has 8gb of ram and runs great. Windows 11 can't compete on that hardware because it's so bloated.
I find it strange given microsoft made windows 3.11, which ran fine on computers with just 4mb of ram. Windows 11 is of course more complex than windows 3.11. But not so much that it should need literally 2000 times as much ram.
In some ways, the memory apocalypse is a great thing. The windows start menu should never have been made in javascript and react. I'm hoping our industry uses this as a push to make our software small and fast again. If ram is 5x more expensive, we should aim to make 2gb of ram be more than enough for regular computing tasks.
I suspect if we graphed them, they’d pretty perfectly match the computers that were available. Ie, I think windows is optimised just enough to run on the low end hardware of each generation.
https://www.bbc.com/news/technology-67911517
To be more precise, the phones weren’t slowed down to keep the batteries from failing, they were slowed down to maintain stability and stop random crashes. As batteries age they lose their peak voltage capability, so you’re basically undervolting your components unless you slow them down.
As an aside, I have this lovely tower from France, barely used, scrap price should be very high. Priced to sell?
When I build a PC, I pay a little extra because I want it to last a long time. I aim for 5 years, and in reality it usually ends up 7 years (life happens, need the money, etc).
I am now in year 8 of my current PC. I only had money to replace it this year, and am now stuck with it due to the RAM prices. It's going to really suck for me if my RAM fails in the next few years.
The cost of the prebuilt is now lower than today 5090 price.
The PS5/Xbox Series console generation is supposed to be wrapping up soon even though it barely feels like it got started. I don't sense there's appetite for more powerful consoles anytime soon (I don't believe the PS5 Pro made waves like the PS4 Pro did), and jumps in console performance are what determine when games start targeting higher system requirements. So I like to think I can get a good number of years of use out of this.
The problem is I also operated under the prior "storage always gets cheaper over time" and was right at about 80% full when all this insanity hit. I'm now at 90% on my largest ZFS array (~280TB) of spinners... and I get to decide if I want to drop a few paychecks on the next vdev upgrade, or start deleting things.
Not to mention I've been holding off replacing a bunch of old 1TB SATA drives (19 years old now!!) as my scratch partition, as I was going to wait until 8TB SSD got really cheap and just buy two of them for a RAID1 for the use. And now I need those drive slots to upgrade the mass storage pool.
Talk about backing oneself into a corner at the wrong time!
I also had another order where I paid £30 for a 1TB SSD (sata though) and it is now more than £100...
The possibility that these fail does concern me.
They should each buy out a RAM manufacture, and make this the new normal.
It really is the best thing for AI adoption and the shareholders.
It's a totally broken way to practice economics, but alas, it's ours until we change it.
Until the penalties involve actual jail time or a fine 10x larger than the profit they made from doing it, they couldn't care less.
The problem is... now for anyone looking to pickup a slightly older used/surplus server rack - you have the go and buy the (now marked up) RAM separate.
So... did they list the "0GB" racks with removed ram cheaper you ask? Nope. Many of them seem to be basically the same price as they were or even higher than they were, except now you're paying for RAM separate.
I bought a surplus Gigabyte server about 2 years ago off ebay for a home project, and I just went and the same seller still has some for sale, even older now than they were when I grabbed one, but they are 2x the price, and they don't come with any RAM, whereas I got one already populated as they were from the factory. Once you add that back in literally looking at 150-200% markup in the last 2 years on what was just surplus computer equipment that is 5+ years old.
Could this be an indication that there's a supply shortage/price increase of the remaining components in the rack? Or do you interpret it as people just trying their luck out of sheer greed?
I honestly don't think they need to fix their prices, and it's possible that selling to consumers is a secondary market for them compared to datacenters, like Nvidia selling GPUs to gamers. We know that's not where the money is.
Seeing some companies actually exit the consumer market is evidence for this.
When parent commenter says "They have done this in the past after all" that's exactly what they're talking about. Micron, Samsung, Hyinx (prior to SK Group acquisition), and others all participated in RAM price fixing in the early 2000's.
https://en.wikipedia.org/wiki/DRAM_industry_price_fixing
Over the long term demand will rationalise and supply will adjust. But in between there is a squeeze, and the right price is the maximum price buyers are willing to bid for the limited supply. Sucks for other buyers, on the other hand it ensures demand destruction, i.e. that the limited supply goes to the buyer who need it the most, other buyers will delay their purchase and do with what they have.
They have slow rolled rollouts and everything, if there wasn't CXMT they would probably still wouldn't have budged on building new fabs for a while longer.
You can absolutely price fix/manufacture a shortage. Ever heard of merchants setting fire to crops after hoarding to sell at a higher price?
The issue is the distributer and producer are the very same people this essentially means producers have no incentive to rationalize supply, they can make much more money of the long drawn out shortage.
If you want evidence that no one predicted this, look at WD. They disinvested from sandisk up to early this year. Now sandisk is worth more than WD. Doesn't seem consistent with a concerted strategy to gauge prices.
> You can absolutely price fix/manufacture a shortage. Ever heard of merchants setting fire to crops after hoarding to sell at a higher price?
Fine. Show me which manufacturer is cutting production to corner the market. I don't see that.
https://www.blocksandfiles.com/ai-ml/2026/01/20/samsung-sk-h...
Big manufactures have absolutely cut production of lower margin products and it's not like they can re-purpose all of the equipment used for it for HBM yet. edit They can definitely focus more resource because of it on HBM though.
Things have improved a bit since govts are starting to get angry.
Tldr; no one is cutting the production of high margin stuff, but low margin stuff can go and die, because it's not a free market it's a stock market these companies want to show off better margins in every unit there is no reason to keep inefficient things running when not doing it pays more.
They just don't want 30-40% margins anymore. That is not free market that is quite literally serving their own interests because they know there is a monopoly.
All big memory manufacturers are heavily supported by the govt, that's what has made them the monopolies they are Micron supported by USA (they crushed Japan's memory industry for it), and Korean giants are more tangled up with the govt than well most people in US would realize.
Somehow people think free market means anything goes, which honestly is where we stand today so I don't blame anyone.
But using govts to crush competitions and build yourself up on subsidies and similar isn't really free markets. China hasn't been able to grow a significant memory chip industry yet because of lack of EUV machines and similar technologies blocked by the US, which is also not "free market".
As for WD, and Sandisk separating I feel like this was more mutual and they new HDDs or SSDs being together was a distraction for both. Which feels more strategic, think of it like this, the shareholders who had the shares how hold shares in both and together they might have likely been worth less so shareholders made money.
Aka good for free market. I don't see how that relates to the actual memory crunch issue more broadly.
...that's how the market works. This is called capital discipline and is standard across industries. If the increase in demand is temporary, they would be spending 5-10 years of profit on facilities which will only lose money. The same works in non-capitalist economies too, you don't want to waste resources on stranded assets and many still believe this RAM demand is temporary and the house of cards will collapse.
That's EXACTLY how a free market is supposed to work. The existing memory manufacturers are FREE to not build new factories, and anyone else is FREE to build those factories in their place.
I don't know if I would personally fund such a venture though, for roughly the same reason existing manufacturers aren't expanding production.
The question is how quickly the supply adjusts. In a free market, the suppliers would respond to the price signal by quickly expanding their production to meet the elevated demand.
But a market captured by a cartel will be much slower to expand the supply.
So, the test isn't "how much is price going up" it's "how aggressively are the RAM producers expanding supply production"
Cartels aren't known for massively increasing production to satisfy all demand. Far from it I'd say, they're more likely to expand supply to such a level that they can still reap massive profits. Very few countries/companies have the knowhow and willingness to be able to build more.
OPEC is a perfect example of this behaviour. They set hearly production goals to keep prices high. Or the Quebec Maple Syrup Producers organization that limits the amount of syrup a farmer can produce and sell in a year to keep prices high.
Well guess what :) They've done it before, they will do it again.
"To date, five manufacturers have pleaded guilty to their involvement in an international price-fixing conspiracy between July 1, 1998, and June 15, 2002, including Hynix, Infineon, Micron Technology, Samsung, and Elpida.[1]"
[1]https://en.wikipedia.org/wiki/DRAM_industry_price_fixing
And the reason the market is so small is because of players who did as you propose, the DRAM market crashed on them (not due to them but due to broader market dynamics), and they went out of business.
So there is, quite perversely, strong incentive to collude if possible and if not to abide by an unstated agreement to be extremely slow to expand DRAM capacity.
That said...
SK Hynix has two fabs under construction, one set to open later this year. One for 80k wpm, another 360k wpm. A third was recently approved to start construction.
Micro has an overhaul of Fab 6 in Virgina near done and slated to produce first wafers this year. It also has a large expansion in Boise nearly completion for 300k wpm. They're also building a "megafab" site in New York (my guess is >1m wpm) and expanding at their Japanese plants (40k wpm).
Samsung has two under construction, P4 for 100k wpm.
Nanya - historically a secondary player focused only on producing non-leading-edge DRAM is making a play for the leading-edge market with a new fab dedicated to DDR5 and HBM.
It appears they've all decided that the demand is sufficiently sustained that building new fabs is worth doing... but it takes time. If my guess is right Micron is making the largest play with their New York site at 1-1.5m wafers per month. That would give them a large chunk of the market.
(wpm = wafers per month)
Don't worry they will leave money on the table and lobby for protectionism when china comes and eats their lunch.
It really is. First thing I searched for in the thread.
I think we're in agreement here. I don't think memory manufacturers are interested in a market disruption even though profits are incredibly high for a few years. Make it too profitable for too long and others will arise but keep the supply at a steady pace with decent margins and they will remain the only ones on the block because of massive upfront cost.
China is, to my knowledge, the only country that has so far had the willingness to seriously invest. Not necessarily for profits but for independence reasons. I don't see many (or any really) other countries following suit.
Can you just not join the organization?
Maple trees are living things and have a right to reproduce, you can't fault a bird for shitting maple seeds all over my farm planting maple trees everywhere.
If only building a ram FAB was that easy :).
The best hope is that cxmt gets going quicker than expected and is able to satisfy a decent chunk of the Chinese demand for RAM which would - in theory - reduce the demand for ram from the few other producers.
> Backed by the Canadian civil courts, the federation has the monopoly for selling Quebecois maple syrup on the wholesale market, and for exporting it outside the province. Producers are only allowed to sell independently cans of less than five litres or 5kg to visitors to their farm. https://www.bbc.com/news/business-35028380
Lot of unsubstantiated claims with low fidelity meme evidence
https://stream-ai.s3.us-west-1.amazonaws.com/class-action-co...
I do see people arguing, "they had years to increase production - why didn't they?" They've been remarkably candid about this: because RAM production has been losing them money in the past. Because they don't have a cartel, prices were on the floor. Investing in more production would have been corporate suicide.
Even now, there is considerable risk that the bubble will burst, and all this investment in new production will be flushed down the toilet. RAM production is a brutal business.
I said:
> the test isn't "how much is price going up" it's "how aggressively are the RAM producers expanding supply production"
Even when it comes to manufacturing the tooling/testing infrastructure, it's just a question of money aiui.
I didn't say that there was a way to do it in the short run. I laid out what the test is to determine if there's cartel-like behavior or competitive-market-like behavior.
I will say, in a market with only 3 players and very high cap-ex cost, my prior is that you're likely to see more cartel-like behavior than competitive-market-like behavior, but that's still to be determined.
Oh yea, governments can fix by putting 100 billion in memory fabs right now, and when demand crater in decade or so, let tax payer eat the cost of those fabs. At least it will benefit poor computer users like me.
Why would they?
And you claim to have historic examples where they did?
Now what do you think happens when developing new supply is substantially constrained by lead times?
Coz a whole lot less of you have ever dealt with manufacturing logistics.
Some supply and demand curves move in days, some in months, some on years.
That said, even though we might have to wait years for new supply to hit there market, we probably don't have to wait years to see signals of that. We will likely know that manufacturers are starting to build new supply long before we see the chips hit the market.
Even if you started a new fab today, it's 5 years till it comes online. The great hope here is that China's DDR5 yields represent a high reject rate of an enormous amount of new capacity that they'll dump into the market to get a sovereign capability in that area.
This system is too complex, the few individual players too big, with too many artificial distortions for the adjustment to equalize out in a way that's anything but years-long (it may have shock-waves that extend past the decade), and highly destructive to everyone who isn't OPEC, or their largest customers, in this analogy.
The brunt of it will be paid by the people further down the food chain.
And given that these are independent actors in the free market, they’ve seen the history of boom/bust cycles in this market
Free market doesn’t mean “everything just magically happens”
That’s a good one!
When you don't understand how anything works, everything looks like a conspiracy.
20 years ago: https://en.wikipedia.org/wiki/DRAM_industry_price_fixing
> According to the one-count charge filed in San Francisco's federal court on Thursday, Park conspired with unnamed employees from other memory makers to fix the price of DRAM sold to computer makers from April 1, 2001, to June 15, 2002. The government says the move directly affected sales to U.S. computer makers Dell, Hewlett-Packard, Compaq, IBM, Apple Computer, and Gateway.
etc
It's pretty simple. In a correctly regulated market the sellers are in competition with each other so there is a limit to how much they can increase prices, since they'll just lose business to their competitor.
In a poorly regulated market that allows competitors to collude with each other to raise prices by the same amount, they'll inflate prices much further.
No? It goes to the buyer with the deepest pockets.
I'm sure it will only suck for the buyers. This has no effect on inflation and probably will be overlooked by the current admin.
The question is how much of this demand is actually based on AI, or just an excuse to fleece regular buyers (and AI is not really the actual demand driving such prices).
Memory manufacturers have driven up prices via collusion many times in the past (and paid fines for it), without AI or some similar high demand vertical market existing.
But I don't think it's the case now, see my answer elsewhere here for why.
In a market with healthy competition (no collusion), this generally can't happen. Every single vendor is incentivized to lower their prices slightly at all times, with a floor at cost plus margin.
We're talking about the memory market here.
The "healthy competition (no collusion). Every single vendor is incentivized to lower their prices slightly at all times, with a floor at cost plus margin" is the fairy-tale version they teach at 101s.
Everything else you said is accurate, but this is not. Supply will go to those most able to pay, not those who need it the most. It is not a moral allocation, which your language implies.
The actual deals were hours apart and secret for that reason.
It's hilarious watching the conspiracy theorists do their usual thing when there was an actual conspiracy, it was perpetuated by the inner circle of a US company and announced and publicized in full once it had achieved its goal.
Ironically enough it's because of its bet on RAM companies that all of Situational Awareness' publicly trading stocks got bought by the highest bidder (Citadel) when the (leveraged) shit hit the proverbial fan and Situational Awareness created a hole of tens of billions.
Situational Awareness was not forced to sell its share in Anthropic for, by contract, investors in Situational Awareness have no right to force the sale (before this month of September I think) of shares in companies not publicly trading yet.
Citadel bought those RAM stocks right at the end of July, during the dip. Amazing timing for for example MU bounced back from $746 back to above $1000 (now around $950).
Now Situational Awareness may still do exceptionally well: they've got $5 billion or so left in Anthropic AIUI.
In addition to that there are extra fabs build and (for example) China can use the high margin to catch up with their processes without breaking the bank.
An egg takes 21 days to hatch.
Then it's 18-24 weeks to start laying their own eggs. That's ~$5-15 in feed over the period, maybe 40-50 pounds. There's utilities, labor, land, etc., but your incremental cost to add a hen if you have the space is minimal. My numbers won't reflect a lot of operators, as scale matters.
Out of a typical 1.5-3 year lifespan that had a startup cost of a few dollars per, paying back its return by about month 6-7, that's not too bad. It's only worth doing financially in this economy at ludicrous, arguably inherently inhumane scales, but that's what it is on paper.
Not quite the same path to course-correction as a silicon shortage.
"Manufacturers taking advantage of it" wouldn't take it to 500% without AI, because it's way beyond the price elasticity points where they make up lost sales at higher prices.
Those prices lose most consumer sales, and are only viable if there's constant high demand that's relatively insensitive to price (so, not regular folks, but e.g. AI companies).
Eventually someone will manufacture more to undercut
See the tariff thing where Bessent bought rights to refunds
It's all a really low tech con perpetuated by weak elders who need blue pills to simulate virility. Odd how they complain about the kids virtual reality.
Oh no it's not either; it's intentional petty bullying
While they have spiked this year, the trend is towards more RAM for cheaper.
https://ourworldindata.org/grapher/historical-cost-of-comput...
I've always been fine with being "behind" on the latest and greatest. I don't mind watching a movie, playing a game, buying a cell phone, etc, that was released a few years prior. If it was really all that great when it came out it should still be good later on.
Buying slow usually saves a.good bit of money when prices go down over time... but this time they went way up and now I'm kicking myself on it. Not upgrading the computer for now, but it is a bit overdue since I had already put it off for many years.
My main desktop computer is Haswell era. Still does fine, for what I do.
R9700 will lose less than 30% of value in the upcoming 3 years. Give or take. Yet people buy ancient cards for way more than 30% of R9700's cost.
When taking the resale value of used kit on the second-hand market into account, one must also take into account the hassle of selling on the second-hand market.
For me? Personally? That hassle absolutely eliminates any money I would get from any plausible sale price. If I'm getting rid of used kit, it's going to be as donations to people I personally know who will come and pick it up at the agreed upon time.
What I have is still fine, though I do regret cancelling now.
"Due to the continued rise in the prices of core components globally, the production cost of monitors has increased significantly, and we can no longer absorb this pressure on our own."
https://respawnfirst.com/we-can-no-longer-absorb-this-ktc-wa...
Also, when OpenAI's new device comes out, would people shift to talking to these small devices?
Open source https://github.com/cnygaard/glq
Pypi glq https://pypi.org/project/glq/
I decided to get the extra RAM. Boy am I glad now. If I thought this was just a temporary spike, I would be tempted to sell the extra memory (I don't really need it yet), but I don't see prices falling anytime soon.
If Laszlo Hanyecz knew that, after 10 years, 1 BTC will exceed $50K, he wouldn't trade his 10,000 Bitcoins (~$41) for two pizzas!
Here's hoping I can wait out the current shortage before any upgrades (I typically only upgrade every 5-7 years or so anyway) or replacements, because I do not plan to pay the new price.