What's happening right now is very different than what happened back during the dot com crash. When they were doing price fixing it was because there was a glut of supply and demand was tanking. Prices were falling and they coordinated to keep prices from falling less.
Right now demand for DRAM is extremely high bordering on endless. Prices are going up. The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
It would also be dumb to cut production when prices are high because you increase the incentive for one of the outside players to suddenly ramp up production and jump in the market.
Not saying they aren't coordinating in other ways (following each other's leads on price hikes and availability). But again the context here is literally the opposite as last time.
> The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
I would argue that the DRAM price fixing scandal actually demonstrates that the industry operates like a cartel. During times of low demand and high supply, they will coordinate to protect prices, and then during spikes in demand (or alleged spikes in demand) they coordinate to keep the price from dropping.
> then during spikes in demand (or alleged spikes in demand) they coordinate to keep the price from dropping.
Why would they need to coordinate to keep the price from dropping during a spike in demand? a spike in demand will obviously not be expected to lower prices regardless of collusion
Yes, the industry is capacity limited so if there's a true spike in demand, prices will be high even absent any collusion. Especially if previous investment in expanding capacity has been lacking for many years.
If the industry is at capacity (which it plausibly is, especially since HBM memory is made in the same facilities) then no one can physically "undercut" anyone else. Collusion works by artificially restraining supply of some valuable good; if there was genuine collusion at play, we'd probably be seeing companies make less of the expensive HBM (to push its price even higher; note that patent and other IPR restrictions can in fact have this effect, to some extent) and more of the comparatively cheap DRAM!
>Why would they need to coordinate to keep the price from dropping during a spike in demand?
They wouldn't you're right.
But I would expect for them to follow the sorts of behavior we've observed in other markets - egg prices, gasoline prices. When a spike occurs, even if as brief as a lightning strike, they will only very slowly drop prices, when in a purely capitalistic world the price drop ought to be equally fast - suggestive that the slow drop is a mutual agreed upon collusion. After all, it's in all sellers best interest to game that "consumers temporarily agreeable to scalping prices" as hard as possible, Nash equilibrium or whatever amongst sellers. Many such cases and more vicious and brutal punishments for such behavior would serve to benefit the common man, the final point and benefit of capitalism
Of course. Price drops only really come through in response to competition.
Long term high prices invite further investment. Investment arrives and wants a quick return. Fastest way to return is to sell above cost but below market. Established players respond. Yada yada. I once met a former projector salesman who was unbelievably angry that someone, I think Acer, came along and destroyed the ~2000 AUD hard price floor that projectors once commanded, which dropped the whole market and his commissions along with it.
Even when collusion is government endorsed instead of outlawed, the same rules apply. See the bromkonvention. You need the new player willing to take the 10% margin to hurt the bottom line of the guy taking 150%.
>when in a purely capitalistic world the price drop ought to be equally fast
No the price can only drop as fast as supply and competition can catch up. For an industry with high input capex costs, thats extremely slowly. I would think some banks would be keen to take a risk on a new RAM fab based on the demands coming from AI, but also I would personally not take the bet that AI will be in this state in 5 years time. So assuming Banks and other lenders are as skeptical as I am, they wouldn't lend, or would request a bigger entity guarantee the loan.
>brutal punishments for such behavior
Brutal punishments for failing to ramp up production? Or for not lowering prices for no reason? I really dont understand
> when in a purely capitalistic world the price drop ought to be equally fast
That is not obvious. When demand is higher than supply, it is clearly good move to raise prices. But when demand is lower than supply it us not clear than lowering prices would raise volumes to compensate for lower margins.
> Following the plea agreement he was sentenced to 8 months in prison and fined US$250,000.[6] Lee was subsequently promoted to President of Samsung Germany in 2009, and then President of Samsung Europe in 2014.
Was in the room working for a hardware company during the offshoring phase of high tech manufacturing
Offshoring was 100% because of antitrust concerns. Copyright landlords and hardware manufacturers were concerned with Americans and their morals also having the skills to create endless competition.
American workers compensation prize was endless hustle jobs to distract them from political revolution.
The olds don't care if the kids end up unskilled knowledge serfs, fuzzy VHS tapes of outdated academics. The olds will be dead.
The problem with citing cartels and greed is...when prices are low, is it because the industry is momentarily altruistic? Did the industry wake up in late 2025 and think "holy shit, I've been being nice, but it's time to turn over a new leaf and start gouging people?"
I mean, don't get me wrong, they are greedy. But that's been true for years. What's changed is the market.
I do not have any particular insight into the market, I was just extremely confused reading this in particular
> Right now demand for DRAM is extremely high bordering on endless. Prices are going up. The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
Given this first statement, the last one makes no sense.
Either the demand is limited, giving sellers that undercut the ability to move more products... Or the demand is higher then production, making it nonsensical to undercut anyone, because you'll sell out anyway, no matter if you're cheapest or not
If demand is high and supply is fixed (generally the case) you can also pull a DeBeers, grab everyone by the balls, and charge 3x (or more) the already high prices everyone would already get when you sell out your inventory.
Fundamentally because demand and supply curves are lagged. This drives DRAM flood and drought cycles. Right now high DRAM prices are driving fab investment up and demand down. In 3-5 years new fabs will cause DRAM glut that will drive prices down. Lower prices will stimulate demand via things like doubling DRAM size in consumer electronics as companies compete on getting the number bigger. Eventually demand curve will eclipse the supply and we will end up in the DRAM drought again.
This time things are further complicated by the fact that the world is investing a sizable chunk of GDP into building RAM hungry data centers in hopes of building a god which will convert the rest of the world into data centers.
The biggest issue here is that it hurts smaller consumers. Scaling up a fab to produce ram takes 5+ years, people generally buy new hardware ~5 years so this lifecycle of hardware for some people is now locked out. Let's say you was due for an upgrade this year you now might be priced out of the market for the next 5 years and basically due to poor business practices
There is a similar issue happening in the manufacturing space where metal foundries are basically "full" up on allocation for other customers and will refuse to sell to you unless your purchase order is six digits otherwise you pay a hefty premium which once again drives capital towards larger corporations. Compounded by a stagnant jobs market the means that scarcity is just going up and up and nobody is re-investing to meet consumer demands because the market is poisoned by speculation to the absolute extreme
Do people really still upgrade so often? I mean it made sense pre 2015 for desktops, and pre 2020 for laptops... But since... Not much has changed from a performance standpoint.
Even gpus hardly advanced since 2022 (4090) and the next generation is at least 1++ years off. Likely 2-3... And it's unclear wherever it will actually be an upgrade or more of the AI shenanigans they released with the 50 generation.
Well, I just updated my 5950x with 128 GB to AMD Strix Halo with the same amount of much faster RAM. It is noticeable faster, but what's better: the whole computer is tiny and sips energy.
I'm very happy I ordered this in the summer, framework delivered it to me early this month. I wonder will these machine just be out of stock now or the price goes up a lot...
This isn't true. It used to be, as a new fab would appreciably add quantity. At 1M wspm in 2015, a new 100k fab at the most modern node would add effectively 20-30% capacity, and usually multiple.players at once, since all had cash.
Now, the relative shrink is tiny, so capacity adds are just wspm, in effect, and that gives 5%.
Put differently, you cannot invest your way out of the shortage, or into meaningful share....so you take profit.
guess I'm interpretting "1million wspm; add 10%; was effectively a 20-30% capacity increase" in 2015.
Not sure where 5% then comes from. Guessing "relative shrink" is referring to process size (5, 4, 2 nm, whatever) not linearly corresponding to density of transistors, etc.
Kind of funny to see manufacturers get screwed by their own opaque pricing policies for once.
All well and good when you're dictating terms with dozens of buyers, but probably not so much when a single buyer is dictating terms to a couple of sellers.
Samsung and SK Hynix. OpenAI made deals to buy almost half of manufactured DRAM in the world. There's speculation that after this was announced, other companies started making their own deals in a panic, further driving demand.
> after this was announced, other companies started making their own deals in a panic, further driving demand.
Oh ffs it's like the toilet paper thing. I was amazed how long that continued despite credible sources saying there is no shortage, just insane demand from the loonies that don't believe it would return to normal instantly if they would just stop buying more and more extras because "see, it's out again!"
This is almost certainly what's going on right now in the retail market. OTOH it's also a semi-rational response to volatility and uncertainty as to future wholesale prices, due to, e.g. the projected build-out of future AI datacenters. As with any durable good, whenever the price might be expected to rise in the future, people will want to hoard stockpiles and the expected price rise will be brought forward to the present.
Data Center projects that were announced a couple months ago are now beginning to be built out.
Additionally, there have been some supply chain issues the past few years due to trade wars between the US, SK, and China [0], along with the earthquake that hit Taiwan last year [1].
Generally, you feel the pain of supply chain issues within 6-18 months of the initial incident, which is where we are now at because stockpiles have been reduced significantly.
It didn't triple overnight. Contracts for 2026-2027+ hyperscaler orders get negotiated gradually over time and when those contracts are N% higher than last year, ~all supply is spoken for.
Maybe Some nerds joked at lunch that if we hoard all the RAM we cut it off from competitors. They saw what COVID did to supply chains and thought they’d be so smart if they could simulate it.
People pulling their heads out of their ass as to how to actually deploy these systems at scale (AKA to do this effectively, you need to do more than just throw pallets of GPU's at it, such as properly considering Topologies of both NVMe-over-Fabric and PCIe roots/lanes [0]) combined with advances in various technologies (eg RDMA, CXL, cuDF/BaM/GPUD2S/etc) that meaningfully enhance how system ram can be integrated and leveraged are a big part of it.
Also we're hitting that 5 years after DDR5 being readily available which means that a lot of existing enterprise hardware that was on DDR4 is going EOL and being replaced with DDR5 which, given many platforms these days have many more channels available than previously, results in more DRAM being bought than was previously used per node and in total. A lot of enterprise was still buying new DDR4 into 2023 as it was a more affordable way to deploy systems with lots of PCIe lanes which was more important than any the costs associated with the performance gain from DDR5 or related CPU's. (Also, early days DDR5 wasn't really any faster than DDR4 with how loose the timing was unless you were willing to pay a BIG premium)
Regarding the hype of the day: AI specifically, part of it is the rise of wrappers and agents and inference in general that can run on CPU's/leverage system ram. These usecases aren't as sensitive to latency as the training side of things as the network latency from the remote user to the datacenter means latency hits due to hitting the CPU ringbus(infinity fabric, QPI, whatever you want to call it) results in a much less significant share over the overall overhead, and the cost/benefit/availability concerns there has also increased the demand for non-GPU AI compute and RAM.
I wouldn't rule out corruption/price fixing (They've done it before) but I have no evidence of this. Wouldn't surprise me, but I don't think this is it (unless this problem persists for several quarters/years)
There's some geopolitics and FOMO (Corporate keeping up with the joneses) and economics that goes into this as well but I can't really speculate on that specifically, that's not really my area of expertise. Suffice to say, it's kind of like a bank run where it's not so much that the demand itself hit the curve of the hockey stick, but it was gradually increasing until it hit a threshold that was starting to cause delays in delivery/deployments. Given how important many companies view being on the cutting edge here, this lead to sudden spike in volume customers willing to pay premiums for early delivery to hit deployment deadlines, artificially inflating demand and further constraining supply, which just fed back into that feedback loop pushing transient demand even higher.
0: Yes NVMe NAND flash is different than DRAM flash, but the systems/clusters that host the NVMe JBOD's tend to use lots of sysRAM for their index/metadata/"superhot" data layer (think memcached, Redis, the MDS nodes for Lustre, etc), and with the advent of CXL and SCM you can deploy even more DRAM to a cluster/fabric than what is strictly presented by the CPU/mobo's memory controllers/channels. This is not driving overall market volume, but is a source of fierce competition for supply at the very "top" of the DRAM/Flash market.
TL;DR: Convergence of a lot of things driving demand.
>People pulling their heads out of their ass as to how to actually deploy these systems at scale (AKA to do this effectively, you need to do more than just throw pallets of GPU's at it)
Yeah something people don't understand is that the models have become so big it can take minutes to load them from SSDs. If you need to restart a CUDA process for whatever reason, you'd rather want to load the model files from RAM. This means for every GB of VRAM, you also need a GB of system RAM. Then there are things like prefix caching and multi user KV caching. Users generally don't do all their requests one after the other in a short window of time. This means you are better off freeing up VRAM after a minute has passed. If the user sends another request, using the system RAM as cache is still more energy, time and VRAM efficient than recalculation. DDR5 based DRAM is incredibly cheap.
>Regarding the hype of the day: AI specifically, part of it is the rise of wrappers and agents and inference in general that can run on CPU's/leverage system ram.
It has more to do with the dominance of mixture of expert models. Due to expert sparsity the required memory bandwidth drops quite significantly. It is possible to run gpt-oss 20B on a computer with 32GB RAM, a segment that used to be reserved for enthusiasts and developers has now become the mainstream amount of RAM on desktop PCs and mini PCs.
Yeah so if I had to summarize, the problem is that DDR4 is EOL, shifting demand to DDR5. AI demand means people want more than 16GiB RAM (there is actually a flood of used 2x8 GB kits in the laptop DDR5 market). DRAM manufacturers switch to supplying AI data centers and stopped resupplying retailers. Retailers are running out of inventory, leading to a sharp rise in prices. Early production of DDR6 will be out in 2026 with consumer availability in 2027, so there is zero incentive to expand production for DDR5.
DDR4 makes it possible since 5 and half years to build 128GB AMD PC using 4x 32GB DDR4-3200 1.2V JEDEC modules. Only since half a year it is at last possible to build AMD PC with 128GB DDR5-5600 RAM. Because Ryzen DDR5 controller cannot operate with 2 sticks sharing the same channel at JEDEC speed/voltage.
DDR4 just cannot be EOL already because of unbuffered ECC! Even today it is not possible to have 128GB with 2 DDR5 sticks on AMD. Only 96GB.
And actually I'd expect 256GB (DDR5-8800 1.1V JEDEC) to be possible with DDR5. We're now only from 4000 to 5600 5(!) years since the specification was published. What if AMD achieves this only few months before DDR5 EOL?..
> DDR6 will be out in 2026 with consumer availability in 2027
I think implying that the DRAM makers are nefarious is nuts. What would you do in their shoes?
You have to think about what’s happening in the context that its 2025. The US is different… there is no functional financial regulator. The CEO of OpenAI is trying to corner the market and be too big to fail. We’re seeing crazy stock swings with Oracle (whose CEO quit after the September stock event), AMD, Intel, etc. Much of the action is fueled by the circular relationships with OpenAI. Add the drama of malevolent players like Elon & POTUS.
If you’re the CEO of Micron, you’re in a tough spot. You have constrained capacity, delays in your new fab buildout that puts it out of play, and a market that is lucrative, but may be a massive Enron type event.
The people making the shovels need to hold tight. There’s no reason to collude. The smart move is to keep on trucking.
Or they don't want to be the ones holding the bag when all the AI companies that ordered RAM from them end up bankrupt and insolvent ?
The stuff they ordered might even be so custom to be useless to be sold by the creditors, leaving the RAM manufacturers hanging dry with unpaid invoices and idle capacity.
So they just do the old cartel thing and wait it out - with the added benefit of selling the same stuff they have been making for twice the price!
I said this last time, YMTC from China is making NAND and finally crossed the 10% market share. I would not be surprised if they have 30% by 2030. The more NAND Fabs reconfigured to DRAM from Samsung or others the more YMTC will grab NAND market share. So yes naturally those profit current NAND and DRAM player enjoys provides an extra cushion for them to build their Fab which is becoming ever more expensive due it its size and machinery required. But also as war chest. And they know it well.
CXMT's DDR5 and LPDDR5 is also slowly gaining market shares, although not at the pace of YMTC due to yield and cost issues.
Both company are close or already at escape velocity. And then there will be a moment like electric car where DRAM and NAND will oversupply. Which is another reason why DRAM manufactures are eager to move to LPDDR6.
YMTC also announced entry into the DRAM market a couple months back. CXMT recently announced DDR5-8000. Sanctions clearly aren't working to slow progress in China's tech sector, but they seem to be great for the profits of US vassal-states
Every 3-4 years RAM prices spike.
There is always an excuse like a fire in a factory.
I believe the truth is 1) we have little amount of suppliers, and 2) supply is very near the limit of what can be sold.
There's also been not one but two price fixing settlements at different times for the same companies. Almost like every few years they get bold enough to try again, and just settle as the cost of doing business
> Samsung Electronics has lowered its target for NAND wafer output this year to around 4.72 million sheets, about 7% down from the previous year's 5.07 million. Kioxia also adjusted its output from 4.80 million last year to 4.69 million this year.. SK hynix and Micron are likewise keeping output conservatively constrained in a bid to benefit from higher prices. SK hynix's NAND output fell about 10%, from 2.01 million sheets last year to around 1.80 million this year. Micron's situation is similar: it is maintaining production at Fab 7 in Singapore—its largest NAND production base—in the low 300,000-sheet range, keeping a conservative supply posture.
China YMTC (sanctioned by US) and CXMT are increasing production capacity.
Can you blame them, though? Between the notorious boom-and-bust cycle of semiconductor industry, and everyone (including much of this forum) thinking that AI is a bubble that will crash any minute, is it really that unreasonable that they're not trying to massively ramp up supply?
Yes, I can absolutely blame companies for illegally coordinating supply reductions to raise their prices on a captured market, and in this case accidentally causing a global supply crisis in doing so.
Especially when those companies were already caught doing it multiple times over and told to stop.
>but also we don’t like to be grabbed by the balls.
They also don't like being "grabbed by the balls" in a few years time when demand inevitably recedes. It's pretty obvious that people here are throwing a temper tantrum because they need to pay more, and companies aren't rushing in to raise supply, even if they themselves admit such actions would be dumb.
I ordered 96GB of memory last Tuesday from Corsair. Two days later when I checked the website again, the exact same memory was being sold for twice what I paid for it.
Just throwing another anecdote out there, my Zen 1 CPU (1700) was flaking out and I was kind of thinking of doing an AM5 build, with an IPMI motherboard and some nice DDR5 memory. DDR5 has a quasi-ECC built in, which is good for a server.
Thing is, I have 128GB of memtest86-passed DDR4 RAM, and while I don't need that much, the idea of spending ~$300 on 32GB was ludicrous. So I have a Ryzen 5700G now, and all is well.
Yep, it doubled in the last 4 months https://www.youtube.com/watch?v=o5Zc-FsUDCM
I upgraded my PC by adding 64GB.. two Fridays ago I sold the 32 GB I took out for the same amount of what I paid for the 64 GB in July... insane
I bought a refurbished laptop with 64gb ddr4 (so-dimm) last week. It was just slightly more expensive than the 32gb variant with same specs. I guess the seller was not yet aware of the high memory prices.
In a week or two I might be able to make a profit by just selling the memory.
There's cheap adapters that allow you to use SO-DIMM memory in desktop DIMM slots. Of course performance will suffer compared to native DIMM sticks, but if you happen to have SO-DIMM laying around that you aren't currently using, this might be a nice use for it.
This conjures up vivid memories of being in the middle 1990s, with my first Pentium build on a Triton (FX) chipset board. The motherboard only accepted 72-pin SIMMs, but those were still pretty rare and expensive for me.
Instead of buying new expensive RAM, I used memory adapters I used to get that going. IIRC, they were branded SimmVerter.
Each adapter allowed a person to install four 30-pin SIMMs (which were very cheap and common) into one 72-pin SIMM socket on a motherboard.
I had four such adapters, and they were each of different shape: One for each variation of tall-vs-short, and also for left-vs-right. These shapes allowed for all four of the adapters to be used concurrently on a motherboard with 4 72-pin slots without them physically interfering with eachother.
It was so much fun back then to try to get all sixteen 30-pin SIMMs to function reliably.
Things became even more fun when I decided to mix in some 72-pin EDO SIMMs, and get each of them working with their respective features enabled.
(And I said all of that with a lot of sarcasm, but: I did eventually get it all to work reliably-enough for my purposes back then. I also learned a ton of stuff about cleaning electrical contacts, maintaining fast mechanical alignment, conclusively identifying problematic modules through testing, and optimizing physical layouts. These are skills that I've been able to get out and productively use on occasion during the 30-ish years that have subsequently passed.
While I can look back on that with a bit of positivity, I'm not sure that people are ready for that kind of thing today. A lot of folks weren't ready for it back then, either -- adapters like that, while being a [mostly] electrically-sound concept, were popularly considered to be cursed.)
Bought a 64gb upgrade kit in September for my wife's new PC for $205. The same kit right now on Newegg is $570. That's not even double in 4 months; thats almost triple in 2 months.
If I can get that right now $389 retail is there an arb opportunity (not in US, but maybe you are getting fucked on tariffs? Could that be the difference?)
I think it's gonna be like with the price spikes around the time of COVID: corpos complained about the global supply chain disruptions and when all of that was over... just left the prices high up.
It's gonna be the same as what you see with food prices nowadays, even fast food - they just increased a whole bunch and nobody can really do anything about it. It's not like there's that many widespread food chains popping up to replace McDonald's, given that most new players in the market are more likely to go bankrupt than anything.
And when it comes to stuff that's actually hard to manufacture, like GPUs and RAM and other components? The corporations will just lie, or in some cases set the MSRP pretty low (e.g. Intel Arc release) and watch the retailers crank up the prices anyways. Even the one case where it did matter a lot was the Intel Core Ultra release where they didn't move units, but only because of AMD CPUs being cheaper, even those prices were known before release. But even so, who's to say that you're not picking the least bad option which still happens to be bad, how close to BOM pricing are those CPUs? And in that BOM, how many middle men are also pushing their own markup?
What are you gonna do about it, since you can't ever look at their books and validate their claims NOR really buy the product elsewhere because like 3 or 4 corps have an oligopoly over the entire planet's supply? Nothing.
Are we not in a trade war according to some? Prove it by checking their production quotas, if they’re purposely running under capacity then just cancel their patents.
Wright Brothers didn’t make much money from their invention due to the Great War and their patents being canceled to help with the war efforts.
I am currently trying to sell a brand new DDR5 6000 64GB CL 28-36-36-96 kit for 100€ below market prices with warranty (I purchased it 3 months ago, and never opened it as I figured I don't need 128GB in my PC.)
But it's just not selling. I guess most people don't even check ebay, and go straight to hardware online retailers.
personally, i always check ebay. but i know general sentiment around "used/like new" items is fairly poor. just ask grand pap sitting in florida, buying half the computer new, because "he would never buy used", aka some returned tower.
The question is how much of this consumer memory is selling either. It may well be that not many people are actually buying DRAM right now, given how much prices have spiked.
I also have a lot more luck selling PC hardware on bespoke forums than on ebay etc.
Isn't this just the normal process of market clearing, if they are still selling out? It may be a bit coarse grained due to long lead times for more supply but still.
HBM stacks multiple DRAM dies vertically on top of each other (typically 4-12 layers). It’s very different from DDR, especially during manufacturing because it has to be integrated with custom chips. Afaik, this isn’t available to Chinese foundries.. yet.
Thus, they have DDR available. Specifically, theyre using LPDDR5 so far.
Busy getting my new build together so today I bought a 8 Tb WD Black 2280 SSD, a 2 Tb WD Black 2230 SSD and 2 x 64 Gb Crucial SODIMMs for (in comparison) a whopping 1850 USD...
Yes, that's pretty much a complete PC. Or at least it used to be.
Hopefully prices for Radeons remain a bit stable for the coming weeks, I'm still figuring out which one to buy to replace my aging Geforce 1080 and drive my 3840x2160 widescreen...
I am loving my sapphire 9070xt (pulse). It's pretty decent, and gets me ~100 tokens per second on gpt-oss-20b at i think 5bit quant. Bought it for a high high price of 70_000 inr in september
There's no way that Apple pays spot. They will have contracts directly with the manufacturers, guaranteeing them a given supply at a given price. Their whole business kinda depends on it.
Sure, but eventually the contracts will be renewed at possibly a much higher price. It depends on the details of the agreement whether the price is adjusted based on market price periodically too.
If I have a potato field, the cost to me of a potato is whatever equipment and labor it takes to farm one, divided by the yield of potatoes.
If there's suddenly a potato famine, then the cost to other people who don't have a potato field is however much they are willing to pay to avoid starving.
Apple doesn't have a potato field, but what they do have is a ton of negotiating power, market share, and enduring relationships with the manufacturers.
There's also the implicit threat when negotiating with Apple that they might enter your market if you fuck around with them. If Apple perceived DRAM/NAND prices to be a significant threat to their carefully curated pricing structure, they might decide that they need to vertically integrate.
>Memory is being allocated to those with deeper pockets first
Yes, sellers fill the most expensive bids first.
>OpenAI's new "Stargate" project reportedly signed deals with Samsung and SK Hynix for up to 900,000 wafers of DRAM per month to feed its AI clusters, which is an amount close to 40% of total global DRAM output if it's ever met.
Right, so demand has pretty much doubled.
>In this cycle, manufacturers had cut back production and investment during the last downturn (2022), and they've been slow to ramp back up.
And manufacturers have been burnt by over production before and are cautious
>Another factor limiting supply of standard RAM is that memory firms are diverting their limited manufacturing capacity to the most lucrative products. Specifically, there's a gold rush for HBM, which is a special kind of memory used by AI accelerator GPUs, because HBM commands far higher prices and profit margins than commodity DRAM.
See above about filling the more lucrative bids first.
>I'm not saying that all of these reasons given aren't the cause for the recent price boom, but what I am saying is that it wouldn't be the first time that price-fixing occurred in the memory industry.
Ok so what should be done about it?
>While that second lawsuit didn't hold up in court (and failed in appeal), that ongoing suspicion exists for a reason.
K so there have been 2 claims, one spurious.
>Each firm knows that flooding the market would hurt all of their profits, so a form of unspoken coordination can occur, and this is next to impossible to prove.
See: Have been burnt by over production before. And Unspoken coordination is hardly an issue. Reacting to private information is bad, reacting to public information is normal and expected.
>It's hard not to see this supposedly coincidental aligned strategy of restraint and wonder if there's something more at play. All of these actions support pricing stability (for those companies) and suggests that no one is "breaking ranks" to grab a larger share by undercutting prices.
You are telling me that you are suspicious of an industry that has expanded, caused a glut that hurt their own businesses and killed their competitors, and are shying away from repeating that event that you outline as a clear mistake.
If theres so much money on the table, you will need another vendor. New RAM vendors have to weigh risk, including all the risk you have outlined, against what is likely a very massive capital outlay.
I don't understand when people blame AI for buying DDR5 DRAM - aren't they mostly interested in HBM? Or is the fab space being diverted to manufacture more HBM than DDR DRAM previously?
> memory suppliers have both the motive and precedent to coordinate behavior, even tacitly, in order to keep prices high. When only a handful of firms control the taps, it doesn't take a formal cartel for them to collectively benefit from constrained supply. Each firm knows that flooding the market would hurt all of their profits, so a form of unspoken coordination can occur, and this is next to impossible to prove. The backdrop of past cartels makes it hard not to be cynical when hearing that "AI demand" is solely to blame for increased prices. Whether or not any collusion is happening now, it's clear that memory companies are profiting immensely from the current crisis. After bleeding financially during the last oversupply downturn, the major DRAM makers are now seeing record-high earnings in the third quarter of 2025 thanks to the price surge, and to put it bluntly, the shortage is great for business.
A chaebol is a large industrial South Korean conglomerate run and controlled by an individual or family. A chaebol often consists of multiple diversified affiliates, controlled by a person or group. Several dozen large South Korean family-controlled corporate groups fall under this definition.
Not pro monopoly as such, but in the recently posted article on UK shipbuilding they mentioned that a major reason for it's decline was it didn't build out capital capacity after being repeatedly burned by cyclical nature of the industry, with some quote even claiming they thought they would be best positioned for the next bust explicitly.
What choice do these companies have, even assuming the demand doesn't crash? They can build out and cause a glut in a few years. They can not build out and then what? Barriers to entry are high but not insurmountable; long term they'll go the way of UK shipbuilding. It might take a while, though. Also one defector building out would have massive advantage, so it doesn't seem like a stable equilibrium. See also constant bickering in the OPEC, an explicit cartel.
I work in the refurb division of an ewaste recycling company. I was tipped off to these insane prices about two months ago, and switched focus from laptops to drives and RAM, and I'm starting to make bank. Roughly a petabyte of hard drives showed up two weeks ago, and I'm almost halfway through preparing all the drives. Hopefully I'll be able to finish and sell them in time to have a very merry Christmas.
I keep hearing about how the supply and demand cycles are “lagged” and that this price spike happens every 3-5 years for purely economical reasons. I feel like I’m being gaslit — I don’t recall any such price spikes in the past of this magnitude. You’re telling me that there’s a totally predictable price cycle and NO ONE has prepared for it? Or else prices are just high temporarily and no one can step in to increase supply? Either way, it seems that there’s not enough competition in this market.
Okay, this will be a bit on a conspiracy theory side, but there was a paper recently describing how to do matrix computations on a RAM stick connected to FPGA, and they've shown it's possible to do it cheaper per flop that GPUs. Except, of course, there's a variety of RAM producers.
It might either be artificial to keep GPU prices where they are, or someone already started building their RAM-based AI datacenter.
I've seen a fair number of articles suggesting that the finances of AI companies have lost touch with reality and that the AI sector is now well within bubble territory.
If AI companies continue to scale up and buy massive amounts of memory as prices spike, how much will that intensify the spike? Could feedback of this nature cause a price shock sufficient to pop the AI bubble earlier than it might have otherwise? How soon might that happen?
Hey, at least a fab didn't have to burn down this time. That's progress.
I've maxed out the RAM on every (consumer mobo) build I've ever done, and have always ended up appreciating it. I did a Ryzen build with 192GB back in January, so despite the questionable signal integrity of DDR5 especially on AM5 (more than one DIMM on each channel means lower speed) this time turned out to be no exception. I also stocked up on 130TB of HDD and 30TB of SSD, as it was clear we were headed towards some kind of economic disaster. (look, I'm avoiding politics!)
But the best RAM purchase I ever did was during the dot com crash glut a few decades ago. I maxed out my Athlon XP with 3x 512MB sticks for $27 each. A gig and a half of RAM. Those were the days.
My main desktop has 256GB, and my "router" has 128GB. Mostly used for software builds (I use NixOS as a source distribution) and VMs (which themselves are mostly used for web browsing, surprise surprise). The desktop had been fine at 112GB because that was the max that coreboot could train, but the coreboot code has been updated and used DDR3 sticks are cheap. The goals of the new 192GB machine are more of that existing usage, plus LLM experimentation and possible VFX rendering/video work. It's speculative of course, but as I've said I've never regretted bulking up on RAM. If it ends up feeling too vacant and echoey, I can always grab another motherboard and processor and make two 96GB machines.
I’m sad the mid to late 2010s alt reality era has come for HN the last 6 months, after 16 years here.
This was my last respite for intellectual argumentation.
Now, every day there’s multiple stories where you have to be caught up on this cinematic universe where AI is fake and doesn’t work and no one uses it and anyone who does is a grifter and or amateur and or embarrassing and any data centers they build will be a waste and they’re probably not even being built and OpenAI is JUST like pets.com so this is basically the web bubble from 1999…so therefore, $X! (In this case, X = RAM supply shortage is fake and actually just coordinated price gouging)
As my MD friend noted wisely a couple weeks ago: it’s noteworthy how this became a culture after LLMs became ubiquitous and user friendly. It was tons of fun and happy times when we were going to reduce # of radiologists, not software engineers.
Not saying that you're wrong, but of course if there is a real AI bubble, then "not letting it pop" only escalates the damage when it does eventually pop. The best outcome all-around is for the market to form accurate expectations quite promptly about the real potential and capabilities of AI, regardless of the immediate consequences.
My own wild guess is that this spike in RAM and storage costs is more of a potential drag on the tech sector as a whole than AI companies specifically. Maybe we'll see some systems being reengineered to cut the waste and the technical debt throughout and be a bit leaner and meaner, since that will be making a real difference to the bottom line.
actually the problem was population boom. FDR raised the rate from $20 to $35/ounce, and by timeNixon took the USA off standard the global population 3x. So gold is no longer scarce because it isn't backing real currency anymore. Its just metal
Might we perhaps see some meaningful investment into long-term alternatives to DRAM as a result of this price spike? Will Intel bring back their Optane persistent memory? What about HP's memristors/ReRAM? Magnetic core memory/MRAM?
This is what I'm hopeful for, but haven't researched enough into whether there are hardware startups in this space as I believe the majors will chase the hype cycle. Optane being canned always felt like a poor, shortsighted decision to me.
What's happening right now is very different than what happened back during the dot com crash. When they were doing price fixing it was because there was a glut of supply and demand was tanking. Prices were falling and they coordinated to keep prices from falling less.
Right now demand for DRAM is extremely high bordering on endless. Prices are going up. The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
It would also be dumb to cut production when prices are high because you increase the incentive for one of the outside players to suddenly ramp up production and jump in the market.
Not saying they aren't coordinating in other ways (following each other's leads on price hikes and availability). But again the context here is literally the opposite as last time.
> The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
I would argue that the DRAM price fixing scandal actually demonstrates that the industry operates like a cartel. During times of low demand and high supply, they will coordinate to protect prices, and then during spikes in demand (or alleged spikes in demand) they coordinate to keep the price from dropping.
https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
> then during spikes in demand (or alleged spikes in demand) they coordinate to keep the price from dropping.
Why would they need to coordinate to keep the price from dropping during a spike in demand? a spike in demand will obviously not be expected to lower prices regardless of collusion
Yes, the industry is capacity limited so if there's a true spike in demand, prices will be high even absent any collusion. Especially if previous investment in expanding capacity has been lacking for many years.
To keep someone from undercutting others’ prices and causing a competitive pricing war.
Realistically, it wouldn’t be a meaningful drop to the consumer. But it’ll affect some executive’s ability to buy another unnecessary trinket.
If the industry is at capacity (which it plausibly is, especially since HBM memory is made in the same facilities) then no one can physically "undercut" anyone else. Collusion works by artificially restraining supply of some valuable good; if there was genuine collusion at play, we'd probably be seeing companies make less of the expensive HBM (to push its price even higher; note that patent and other IPR restrictions can in fact have this effect, to some extent) and more of the comparatively cheap DRAM!
>Why would they need to coordinate to keep the price from dropping during a spike in demand?
They wouldn't you're right.
But I would expect for them to follow the sorts of behavior we've observed in other markets - egg prices, gasoline prices. When a spike occurs, even if as brief as a lightning strike, they will only very slowly drop prices, when in a purely capitalistic world the price drop ought to be equally fast - suggestive that the slow drop is a mutual agreed upon collusion. After all, it's in all sellers best interest to game that "consumers temporarily agreeable to scalping prices" as hard as possible, Nash equilibrium or whatever amongst sellers. Many such cases and more vicious and brutal punishments for such behavior would serve to benefit the common man, the final point and benefit of capitalism
>they will only very slowly drop prices
Of course. Price drops only really come through in response to competition.
Long term high prices invite further investment. Investment arrives and wants a quick return. Fastest way to return is to sell above cost but below market. Established players respond. Yada yada. I once met a former projector salesman who was unbelievably angry that someone, I think Acer, came along and destroyed the ~2000 AUD hard price floor that projectors once commanded, which dropped the whole market and his commissions along with it.
Even when collusion is government endorsed instead of outlawed, the same rules apply. See the bromkonvention. You need the new player willing to take the 10% margin to hurt the bottom line of the guy taking 150%.
>when in a purely capitalistic world the price drop ought to be equally fast
No the price can only drop as fast as supply and competition can catch up. For an industry with high input capex costs, thats extremely slowly. I would think some banks would be keen to take a risk on a new RAM fab based on the demands coming from AI, but also I would personally not take the bet that AI will be in this state in 5 years time. So assuming Banks and other lenders are as skeptical as I am, they wouldn't lend, or would request a bigger entity guarantee the loan.
>brutal punishments for such behavior
Brutal punishments for failing to ramp up production? Or for not lowering prices for no reason? I really dont understand
> when in a purely capitalistic world the price drop ought to be equally fast
That is not obvious. When demand is higher than supply, it is clearly good move to raise prices. But when demand is lower than supply it us not clear than lowering prices would raise volumes to compensate for lower margins.
If there are a small number of sellers and barriers to entry are high, we should expect at the bare minimum "tacit collusion":
https://en.wikipedia.org/wiki/Tacit_collusion
> Following the plea agreement he was sentenced to 8 months in prison and fined US$250,000.[6] Lee was subsequently promoted to President of Samsung Germany in 2009, and then President of Samsung Europe in 2014.
Lmao he got rewarded for taking the fall.
Was in the room working for a hardware company during the offshoring phase of high tech manufacturing
Offshoring was 100% because of antitrust concerns. Copyright landlords and hardware manufacturers were concerned with Americans and their morals also having the skills to create endless competition.
American workers compensation prize was endless hustle jobs to distract them from political revolution.
The olds don't care if the kids end up unskilled knowledge serfs, fuzzy VHS tapes of outdated academics. The olds will be dead.
Said industry's cartel behavior is already a matter of historical fact, so they do not earn any benefit of the doubt being afforded them.
The problem with citing cartels and greed is...when prices are low, is it because the industry is momentarily altruistic? Did the industry wake up in late 2025 and think "holy shit, I've been being nice, but it's time to turn over a new leaf and start gouging people?"
I mean, don't get me wrong, they are greedy. But that's been true for years. What's changed is the market.
i thought samsung took a fab offline as well?
I do not have any particular insight into the market, I was just extremely confused reading this in particular
> Right now demand for DRAM is extremely high bordering on endless. Prices are going up. The incentive for one of the big players to undercut the other on cost even just a little bit to pick up market share is extremely lucrative.
Given this first statement, the last one makes no sense.
Either the demand is limited, giving sellers that undercut the ability to move more products... Or the demand is higher then production, making it nonsensical to undercut anyone, because you'll sell out anyway, no matter if you're cheapest or not
If demand is high and supply is fixed (generally the case) you can also pull a DeBeers, grab everyone by the balls, and charge 3x (or more) the already high prices everyone would already get when you sell out your inventory.
To the extent that demand is price inelastic, which in practice means a lack of substitutes and the product serving an essential need.
With the AI boom ongoing, plenty of people with deep pockets are desperate. Perfect time for some gouging.
Demand for diamonds is rather elastic, yet DeBrees did it.
> Right now demand for DRAM is extremely high bordering on endless
Why is demand extremely high right now when it wasn't a couple of months ago? What changed since then that caused it to triple overnight?
Fundamentally because demand and supply curves are lagged. This drives DRAM flood and drought cycles. Right now high DRAM prices are driving fab investment up and demand down. In 3-5 years new fabs will cause DRAM glut that will drive prices down. Lower prices will stimulate demand via things like doubling DRAM size in consumer electronics as companies compete on getting the number bigger. Eventually demand curve will eclipse the supply and we will end up in the DRAM drought again.
This time things are further complicated by the fact that the world is investing a sizable chunk of GDP into building RAM hungry data centers in hopes of building a god which will convert the rest of the world into data centers.
> in hopes of building a god which will convert the rest of the world into data centers
I'm as pro AI as it comes, and I love your way of putting it. Very poignant.
if its what it takes then so be it
which company achieve AGI first would be a quadrillion company
If you make something that's as smart as a sloppy intern and costs ten thousand dollars an hour to run, you're not going to make any money off of it.
But if you make something that's as smart as a sloppy intern and costs even a small amount less, you're making an unbelievable amount of money off it.
Are you sure you wouldn't just lower the market price for an intern and the intern would still be at least at the same price?
Also an intern gains experience, while it's wage doesn't automatically.
Intern needs breaks and vacation days. AGI don’t
When our main comparison is cost per work hour, that doesn't matter a whole lot.
The biggest issue here is that it hurts smaller consumers. Scaling up a fab to produce ram takes 5+ years, people generally buy new hardware ~5 years so this lifecycle of hardware for some people is now locked out. Let's say you was due for an upgrade this year you now might be priced out of the market for the next 5 years and basically due to poor business practices
There is a similar issue happening in the manufacturing space where metal foundries are basically "full" up on allocation for other customers and will refuse to sell to you unless your purchase order is six digits otherwise you pay a hefty premium which once again drives capital towards larger corporations. Compounded by a stagnant jobs market the means that scarcity is just going up and up and nobody is re-investing to meet consumer demands because the market is poisoned by speculation to the absolute extreme
Do people really still upgrade so often? I mean it made sense pre 2015 for desktops, and pre 2020 for laptops... But since... Not much has changed from a performance standpoint.
Even gpus hardly advanced since 2022 (4090) and the next generation is at least 1++ years off. Likely 2-3... And it's unclear wherever it will actually be an upgrade or more of the AI shenanigans they released with the 50 generation.
I'm still on a 2013 desktop I found in a dumpster, its mostly fine still, apart from ballooning memory use in browsers.
Well, I just updated my 5950x with 128 GB to AMD Strix Halo with the same amount of much faster RAM. It is noticeable faster, but what's better: the whole computer is tiny and sips energy.
I'm very happy I ordered this in the summer, framework delivered it to me early this month. I wonder will these machine just be out of stock now or the price goes up a lot...
Laptops are pretty fragile, especially with how a lot of people treat them. Replacement is due to breakage, not obsolescence.
If you have an old enough Laptop, you can just replace parts.
This isn't true. It used to be, as a new fab would appreciably add quantity. At 1M wspm in 2015, a new 100k fab at the most modern node would add effectively 20-30% capacity, and usually multiple.players at once, since all had cash.
Now, the relative shrink is tiny, so capacity adds are just wspm, in effect, and that gives 5%.
Put differently, you cannot invest your way out of the shortage, or into meaningful share....so you take profit.
apparently wspm is Wafer Starts Per Month
guess I'm interpretting "1million wspm; add 10%; was effectively a 20-30% capacity increase" in 2015.
Not sure where 5% then comes from. Guessing "relative shrink" is referring to process size (5, 4, 2 nm, whatever) not linearly corresponding to density of transistors, etc.
I've heard it was Sam Altman and OpenAI basically buying every wafer available from both Samsung and SKHynix at the start of October.
Neither company know of the other purchase until it was a done deal.
Kind of funny to see manufacturers get screwed by their own opaque pricing policies for once.
All well and good when you're dictating terms with dozens of buyers, but probably not so much when a single buyer is dictating terms to a couple of sellers.
The CEO of OpenAI and OpenAI didn't coordinate?
Samsung and SK Hynix. OpenAI made deals to buy almost half of manufactured DRAM in the world. There's speculation that after this was announced, other companies started making their own deals in a panic, further driving demand.
> after this was announced, other companies started making their own deals in a panic, further driving demand.
Oh ffs it's like the toilet paper thing. I was amazed how long that continued despite credible sources saying there is no shortage, just insane demand from the loonies that don't believe it would return to normal instantly if they would just stop buying more and more extras because "see, it's out again!"
This is almost certainly what's going on right now in the retail market. OTOH it's also a semi-rational response to volatility and uncertainty as to future wholesale prices, due to, e.g. the projected build-out of future AI datacenters. As with any durable good, whenever the price might be expected to rise in the future, people will want to hoard stockpiles and the expected price rise will be brought forward to the present.
Is DRAM built using the same process/node as GPUs?
> Why is demand extremely high...
Data Center projects that were announced a couple months ago are now beginning to be built out.
Additionally, there have been some supply chain issues the past few years due to trade wars between the US, SK, and China [0], along with the earthquake that hit Taiwan last year [1].
Generally, you feel the pain of supply chain issues within 6-18 months of the initial incident, which is where we are now at because stockpiles have been reduced significantly.
[0] - https://www.digitimes.com/news/a20230310PD204/chip-war-memor...
[1] - https://www.reuters.com/technology/micron-flags-hit-its-dram...
It didn't triple overnight. Contracts for 2026-2027+ hyperscaler orders get negotiated gradually over time and when those contracts are N% higher than last year, ~all supply is spoken for.
Maybe Some nerds joked at lunch that if we hoard all the RAM we cut it off from competitors. They saw what COVID did to supply chains and thought they’d be so smart if they could simulate it.
DRAM 2025 as toilet paper from 2020. Seems plausible.
No no no, it’s all just the bullwhip effect, remember? Remember when prices went fucking nuts during Covid and then dropped?
Oh wait, they never fucking dropped.
Still waiting, all you bullshit, er, bullwhip truthers out there.
People pulling their heads out of their ass as to how to actually deploy these systems at scale (AKA to do this effectively, you need to do more than just throw pallets of GPU's at it, such as properly considering Topologies of both NVMe-over-Fabric and PCIe roots/lanes [0]) combined with advances in various technologies (eg RDMA, CXL, cuDF/BaM/GPUD2S/etc) that meaningfully enhance how system ram can be integrated and leveraged are a big part of it.
Also we're hitting that 5 years after DDR5 being readily available which means that a lot of existing enterprise hardware that was on DDR4 is going EOL and being replaced with DDR5 which, given many platforms these days have many more channels available than previously, results in more DRAM being bought than was previously used per node and in total. A lot of enterprise was still buying new DDR4 into 2023 as it was a more affordable way to deploy systems with lots of PCIe lanes which was more important than any the costs associated with the performance gain from DDR5 or related CPU's. (Also, early days DDR5 wasn't really any faster than DDR4 with how loose the timing was unless you were willing to pay a BIG premium)
Regarding the hype of the day: AI specifically, part of it is the rise of wrappers and agents and inference in general that can run on CPU's/leverage system ram. These usecases aren't as sensitive to latency as the training side of things as the network latency from the remote user to the datacenter means latency hits due to hitting the CPU ringbus(infinity fabric, QPI, whatever you want to call it) results in a much less significant share over the overall overhead, and the cost/benefit/availability concerns there has also increased the demand for non-GPU AI compute and RAM.
I wouldn't rule out corruption/price fixing (They've done it before) but I have no evidence of this. Wouldn't surprise me, but I don't think this is it (unless this problem persists for several quarters/years)
There's some geopolitics and FOMO (Corporate keeping up with the joneses) and economics that goes into this as well but I can't really speculate on that specifically, that's not really my area of expertise. Suffice to say, it's kind of like a bank run where it's not so much that the demand itself hit the curve of the hockey stick, but it was gradually increasing until it hit a threshold that was starting to cause delays in delivery/deployments. Given how important many companies view being on the cutting edge here, this lead to sudden spike in volume customers willing to pay premiums for early delivery to hit deployment deadlines, artificially inflating demand and further constraining supply, which just fed back into that feedback loop pushing transient demand even higher.
0: Yes NVMe NAND flash is different than DRAM flash, but the systems/clusters that host the NVMe JBOD's tend to use lots of sysRAM for their index/metadata/"superhot" data layer (think memcached, Redis, the MDS nodes for Lustre, etc), and with the advent of CXL and SCM you can deploy even more DRAM to a cluster/fabric than what is strictly presented by the CPU/mobo's memory controllers/channels. This is not driving overall market volume, but is a source of fierce competition for supply at the very "top" of the DRAM/Flash market.
TL;DR: Convergence of a lot of things driving demand.
>People pulling their heads out of their ass as to how to actually deploy these systems at scale (AKA to do this effectively, you need to do more than just throw pallets of GPU's at it)
Yeah something people don't understand is that the models have become so big it can take minutes to load them from SSDs. If you need to restart a CUDA process for whatever reason, you'd rather want to load the model files from RAM. This means for every GB of VRAM, you also need a GB of system RAM. Then there are things like prefix caching and multi user KV caching. Users generally don't do all their requests one after the other in a short window of time. This means you are better off freeing up VRAM after a minute has passed. If the user sends another request, using the system RAM as cache is still more energy, time and VRAM efficient than recalculation. DDR5 based DRAM is incredibly cheap.
>Regarding the hype of the day: AI specifically, part of it is the rise of wrappers and agents and inference in general that can run on CPU's/leverage system ram.
It has more to do with the dominance of mixture of expert models. Due to expert sparsity the required memory bandwidth drops quite significantly. It is possible to run gpt-oss 20B on a computer with 32GB RAM, a segment that used to be reserved for enthusiasts and developers has now become the mainstream amount of RAM on desktop PCs and mini PCs.
Yeah so if I had to summarize, the problem is that DDR4 is EOL, shifting demand to DDR5. AI demand means people want more than 16GiB RAM (there is actually a flood of used 2x8 GB kits in the laptop DDR5 market). DRAM manufacturers switch to supplying AI data centers and stopped resupplying retailers. Retailers are running out of inventory, leading to a sharp rise in prices. Early production of DDR6 will be out in 2026 with consumer availability in 2027, so there is zero incentive to expand production for DDR5.
> DDR4 is EOL
11 years after specification released.
DDR4 makes it possible since 5 and half years to build 128GB AMD PC using 4x 32GB DDR4-3200 1.2V JEDEC modules. Only since half a year it is at last possible to build AMD PC with 128GB DDR5-5600 RAM. Because Ryzen DDR5 controller cannot operate with 2 sticks sharing the same channel at JEDEC speed/voltage.
DDR4 just cannot be EOL already because of unbuffered ECC! Even today it is not possible to have 128GB with 2 DDR5 sticks on AMD. Only 96GB.
And actually I'd expect 256GB (DDR5-8800 1.1V JEDEC) to be possible with DDR5. We're now only from 4000 to 5600 5(!) years since the specification was published. What if AMD achieves this only few months before DDR5 EOL?..
> DDR6 will be out in 2026 with consumer availability in 2027
... or not at all?..
I think implying that the DRAM makers are nefarious is nuts. What would you do in their shoes?
You have to think about what’s happening in the context that its 2025. The US is different… there is no functional financial regulator. The CEO of OpenAI is trying to corner the market and be too big to fail. We’re seeing crazy stock swings with Oracle (whose CEO quit after the September stock event), AMD, Intel, etc. Much of the action is fueled by the circular relationships with OpenAI. Add the drama of malevolent players like Elon & POTUS.
If you’re the CEO of Micron, you’re in a tough spot. You have constrained capacity, delays in your new fab buildout that puts it out of play, and a market that is lucrative, but may be a massive Enron type event.
The people making the shovels need to hold tight. There’s no reason to collude. The smart move is to keep on trucking.
Or they don't want to be the ones holding the bag when all the AI companies that ordered RAM from them end up bankrupt and insolvent ?
The stuff they ordered might even be so custom to be useless to be sold by the creditors, leaving the RAM manufacturers hanging dry with unpaid invoices and idle capacity.
So they just do the old cartel thing and wait it out - with the added benefit of selling the same stuff they have been making for twice the price!
Supply and demand "theory" being used yet again to justify obvious price fixing.
If fixing, Why now? Why this much?
I said this last time, YMTC from China is making NAND and finally crossed the 10% market share. I would not be surprised if they have 30% by 2030. The more NAND Fabs reconfigured to DRAM from Samsung or others the more YMTC will grab NAND market share. So yes naturally those profit current NAND and DRAM player enjoys provides an extra cushion for them to build their Fab which is becoming ever more expensive due it its size and machinery required. But also as war chest. And they know it well.
CXMT's DDR5 and LPDDR5 is also slowly gaining market shares, although not at the pace of YMTC due to yield and cost issues.
Both company are close or already at escape velocity. And then there will be a moment like electric car where DRAM and NAND will oversupply. Which is another reason why DRAM manufactures are eager to move to LPDDR6.
YMTC also announced entry into the DRAM market a couple months back. CXMT recently announced DDR5-8000. Sanctions clearly aren't working to slow progress in China's tech sector, but they seem to be great for the profits of US vassal-states
Every 3-4 years RAM prices spike. There is always an excuse like a fire in a factory. I believe the truth is 1) we have little amount of suppliers, and 2) supply is very near the limit of what can be sold.
https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
There's also been not one but two price fixing settlements at different times for the same companies. Almost like every few years they get bold enough to try again, and just settle as the cost of doing business
"OMEC" (Organization of Memory Exporting Countries) NAND production quotas? https://x.com/jukanlosreve/status/1988505115339436423
> Samsung Electronics has lowered its target for NAND wafer output this year to around 4.72 million sheets, about 7% down from the previous year's 5.07 million. Kioxia also adjusted its output from 4.80 million last year to 4.69 million this year.. SK hynix and Micron are likewise keeping output conservatively constrained in a bid to benefit from higher prices. SK hynix's NAND output fell about 10%, from 2.01 million sheets last year to around 1.80 million this year. Micron's situation is similar: it is maintaining production at Fab 7 in Singapore—its largest NAND production base—in the low 300,000-sheet range, keeping a conservative supply posture.
China YMTC (sanctioned by US) and CXMT are increasing production capacity.
Can you blame them, though? Between the notorious boom-and-bust cycle of semiconductor industry, and everyone (including much of this forum) thinking that AI is a bubble that will crash any minute, is it really that unreasonable that they're not trying to massively ramp up supply?
Yes, I can absolutely blame companies for illegally coordinating supply reductions to raise their prices on a captured market, and in this case accidentally causing a global supply crisis in doing so.
Especially when those companies were already caught doing it multiple times over and told to stop.
>Yes, I can absolutely blame companies for illegally coordinating supply reductions
There's no evidence that they're coordinating, though.
Memory of previous investigation.
In the meantime, Chinese memory suppliers are ramping supply and winning deals with PC OEMs.
Like drug or oil cartels (OPEC), the reasons are obvious and reasonable, but also we don’t like to be grabbed by the balls.
Both parties can have perfectly rational reasons to both exist and hate each other at the same time.
>but also we don’t like to be grabbed by the balls.
They also don't like being "grabbed by the balls" in a few years time when demand inevitably recedes. It's pretty obvious that people here are throwing a temper tantrum because they need to pay more, and companies aren't rushing in to raise supply, even if they themselves admit such actions would be dumb.
We’re consumers so of course we’re the good guys (/s)
That wikipedia page only shows one settlement. The other got dismissed.
There were also dumping controversies in the 1980s. Who knows that may even be where Trump got his crazy tarrif fetish from.
I ChatGPT'd your claim, from recent years it seems correct.
Here's some basic analysis it did (AI, so take with grain of salt): https://shottr.cc/s/2zv5/SCR-20251127-fm6p.png
2016–2018 price spike
DRAM prices nearly tripled.
Driven by smartphone demand + server build-out + constrained supply expansion.
Resulted in the 2018 antitrust lawsuit accusing Samsung, Micron, SK hynix of coordinated output limits.
2019 oversupply crash
After the spike, memory makers over-invested capacity.
Demand softened; inventories built up.
Prices fell sharply — ~35–40% decline.
2020 mild recovery
COVID WFH / remote boom → PC sales up → DRAM demand rose slightly.
Prices recovered modestly.
2021 rebound
Data-center and cloud expansion returned.
DDR5 ramping begins.
Market in recovery — but not frothy.
2022–2023 bottom
Smartphone shipments declined globally.
PC demand dropped post-COVID.
OEMs had too much inventory → massive price declines.
Some DRAM sold below production cost.
2024–2025 AI/HBM super-cycle
AI training systems (NVIDIA, AMD, etc.) require large HBM arrays.
Fab capacity reallocated from commodity DRAM → HBM.
AI hyperscalers (OpenAI, Microsoft, Meta, Tesla) created enormous DRAM draw.
Commodity DRAM restricted → price spike ≥170% YoY.
I ordered 96GB of memory last Tuesday from Corsair. Two days later when I checked the website again, the exact same memory was being sold for twice what I paid for it.
Just throwing another anecdote out there, my Zen 1 CPU (1700) was flaking out and I was kind of thinking of doing an AM5 build, with an IPMI motherboard and some nice DDR5 memory. DDR5 has a quasi-ECC built in, which is good for a server.
Thing is, I have 128GB of memtest86-passed DDR4 RAM, and while I don't need that much, the idea of spending ~$300 on 32GB was ludicrous. So I have a Ryzen 5700G now, and all is well.
Yep, it doubled in the last 4 months https://www.youtube.com/watch?v=o5Zc-FsUDCM I upgraded my PC by adding 64GB.. two Fridays ago I sold the 32 GB I took out for the same amount of what I paid for the 64 GB in July... insane
And I'm beating myself for not preemptively ordering that 128G kit for $500 a couple of months ago, thinking about upgrading soon.
Last week it went to $1300 and now it's not available anymore.
Guess I'll just skip AM5 and wait for AM6 at this rate...
Same situation. I looked up the previous RAM boom and bust cycles (thanks Gemini) and we are looking at over a year of waiting potentially.
I bought a refurbished laptop with 64gb ddr4 (so-dimm) last week. It was just slightly more expensive than the 32gb variant with same specs. I guess the seller was not yet aware of the high memory prices.
In a week or two I might be able to make a profit by just selling the memory.
There's cheap adapters that allow you to use SO-DIMM memory in desktop DIMM slots. Of course performance will suffer compared to native DIMM sticks, but if you happen to have SO-DIMM laying around that you aren't currently using, this might be a nice use for it.
Oh, man.
This conjures up vivid memories of being in the middle 1990s, with my first Pentium build on a Triton (FX) chipset board. The motherboard only accepted 72-pin SIMMs, but those were still pretty rare and expensive for me.
Instead of buying new expensive RAM, I used memory adapters I used to get that going. IIRC, they were branded SimmVerter.
Each adapter allowed a person to install four 30-pin SIMMs (which were very cheap and common) into one 72-pin SIMM socket on a motherboard.
I had four such adapters, and they were each of different shape: One for each variation of tall-vs-short, and also for left-vs-right. These shapes allowed for all four of the adapters to be used concurrently on a motherboard with 4 72-pin slots without them physically interfering with eachother.
It was so much fun back then to try to get all sixteen 30-pin SIMMs to function reliably.
Things became even more fun when I decided to mix in some 72-pin EDO SIMMs, and get each of them working with their respective features enabled.
(And I said all of that with a lot of sarcasm, but: I did eventually get it all to work reliably-enough for my purposes back then. I also learned a ton of stuff about cleaning electrical contacts, maintaining fast mechanical alignment, conclusively identifying problematic modules through testing, and optimizing physical layouts. These are skills that I've been able to get out and productively use on occasion during the 30-ish years that have subsequently passed.
While I can look back on that with a bit of positivity, I'm not sure that people are ready for that kind of thing today. A lot of folks weren't ready for it back then, either -- adapters like that, while being a [mostly] electrically-sound concept, were popularly considered to be cursed.)
Bought a 64gb upgrade kit in September for my wife's new PC for $205. The same kit right now on Newegg is $570. That's not even double in 4 months; thats almost triple in 2 months.
If I can get that right now $389 retail is there an arb opportunity (not in US, but maybe you are getting fucked on tariffs? Could that be the difference?)
I think it's gonna be like with the price spikes around the time of COVID: corpos complained about the global supply chain disruptions and when all of that was over... just left the prices high up.
It's gonna be the same as what you see with food prices nowadays, even fast food - they just increased a whole bunch and nobody can really do anything about it. It's not like there's that many widespread food chains popping up to replace McDonald's, given that most new players in the market are more likely to go bankrupt than anything.
And when it comes to stuff that's actually hard to manufacture, like GPUs and RAM and other components? The corporations will just lie, or in some cases set the MSRP pretty low (e.g. Intel Arc release) and watch the retailers crank up the prices anyways. Even the one case where it did matter a lot was the Intel Core Ultra release where they didn't move units, but only because of AMD CPUs being cheaper, even those prices were known before release. But even so, who's to say that you're not picking the least bad option which still happens to be bad, how close to BOM pricing are those CPUs? And in that BOM, how many middle men are also pushing their own markup?
What are you gonna do about it, since you can't ever look at their books and validate their claims NOR really buy the product elsewhere because like 3 or 4 corps have an oligopoly over the entire planet's supply? Nothing.
Shorten patents to 18months (the industry's innovation cycle, in the past).
I'll bet you'll see a lot more output, esp. with low margins -- to make the market uninteresting for new players.
My personal opinion:
Trade Mark - as long as dues are paid
Patents a yearly cap of N things and only the most worthy get a 20 year patent, world wide.
Copyright ~ 5 years auto with exponentially more expensive renewals in 5 year blocks.
Are we not in a trade war according to some? Prove it by checking their production quotas, if they’re purposely running under capacity then just cancel their patents.
Wright Brothers didn’t make much money from their invention due to the Great War and their patents being canceled to help with the war efforts.
I am currently trying to sell a brand new DDR5 6000 64GB CL 28-36-36-96 kit for 100€ below market prices with warranty (I purchased it 3 months ago, and never opened it as I figured I don't need 128GB in my PC.)
But it's just not selling. I guess most people don't even check ebay, and go straight to hardware online retailers.
personally, i always check ebay. but i know general sentiment around "used/like new" items is fairly poor. just ask grand pap sitting in florida, buying half the computer new, because "he would never buy used", aka some returned tower.
The question is how much of this consumer memory is selling either. It may well be that not many people are actually buying DRAM right now, given how much prices have spiked.
I also have a lot more luck selling PC hardware on bespoke forums than on ebay etc.
Isn't this just the normal process of market clearing, if they are still selling out? It may be a bit coarse grained due to long lead times for more supply but still.
I expected that this was because China use DDR5 in their new AI chips (Due to not having access to (enough) HBM)
HBM is independent of the DDR revision being used. HBM how the DRAM attaches to the chip. Really not sure what you’re trying to say.
HBM stacks multiple DRAM dies vertically on top of each other (typically 4-12 layers). It’s very different from DDR, especially during manufacturing because it has to be integrated with custom chips. Afaik, this isn’t available to Chinese foundries.. yet.
Thus, they have DDR available. Specifically, theyre using LPDDR5 so far.
But it’s still the same dies- the shortage is in the dies not the packaging, no?
It’s difficult to integrate as HBM and the density is very different. China wouldnt select 30x slower bandwith if they had a choice.
Is there a futures market for DRAM? I feels like there should be...
You can always buy some SKHynix or Micron shares.
That doesn't help me lock in prices when I'm giving quotes on hardware that still needs to be manufactured.
of course there is - https://www.dramexchange.com
Unfortunately got some experience with that...
Busy getting my new build together so today I bought a 8 Tb WD Black 2280 SSD, a 2 Tb WD Black 2230 SSD and 2 x 64 Gb Crucial SODIMMs for (in comparison) a whopping 1850 USD...
Yes, that's pretty much a complete PC. Or at least it used to be.
Hopefully prices for Radeons remain a bit stable for the coming weeks, I'm still figuring out which one to buy to replace my aging Geforce 1080 and drive my 3840x2160 widescreen...
I am loving my sapphire 9070xt (pulse). It's pretty decent, and gets me ~100 tokens per second on gpt-oss-20b at i think 5bit quant. Bought it for a high high price of 70_000 inr in september
Quick, buy a Mac with higher-specced memory while the price is almost reasonable.
I'm wondering if Apple will need to adjust their prices to maintain their margins. Given the scale of price rises I'd say yes.
There's no way that Apple pays spot. They will have contracts directly with the manufacturers, guaranteeing them a given supply at a given price. Their whole business kinda depends on it.
Sure, but eventually the contracts will be renewed at possibly a much higher price. It depends on the details of the agreement whether the price is adjusted based on market price periodically too.
If I have a potato field, the cost to me of a potato is whatever equipment and labor it takes to farm one, divided by the yield of potatoes.
If there's suddenly a potato famine, then the cost to other people who don't have a potato field is however much they are willing to pay to avoid starving.
Apple doesn't have a potato field, but what they do have is a ton of negotiating power, market share, and enduring relationships with the manufacturers.
There's also the implicit threat when negotiating with Apple that they might enter your market if you fuck around with them. If Apple perceived DRAM/NAND prices to be a significant threat to their carefully curated pricing structure, they might decide that they need to vertically integrate.
*only slightly less reasonable than market prices before they adjust it in order to stay at 2x less reasonable
What? Apple's memory upgrades are never reasonable.
Last time I checked going from 16 to 32GB in a Mac Mini was more expensive (or as expensive) than buying two 16GB Mac Minis.
Right now in this environment, they look reasonable. Not that they are
>Memory is being allocated to those with deeper pockets first
Yes, sellers fill the most expensive bids first.
>OpenAI's new "Stargate" project reportedly signed deals with Samsung and SK Hynix for up to 900,000 wafers of DRAM per month to feed its AI clusters, which is an amount close to 40% of total global DRAM output if it's ever met.
Right, so demand has pretty much doubled.
>In this cycle, manufacturers had cut back production and investment during the last downturn (2022), and they've been slow to ramp back up.
And manufacturers have been burnt by over production before and are cautious
>Another factor limiting supply of standard RAM is that memory firms are diverting their limited manufacturing capacity to the most lucrative products. Specifically, there's a gold rush for HBM, which is a special kind of memory used by AI accelerator GPUs, because HBM commands far higher prices and profit margins than commodity DRAM.
See above about filling the more lucrative bids first.
>I'm not saying that all of these reasons given aren't the cause for the recent price boom, but what I am saying is that it wouldn't be the first time that price-fixing occurred in the memory industry.
Ok so what should be done about it?
>While that second lawsuit didn't hold up in court (and failed in appeal), that ongoing suspicion exists for a reason.
K so there have been 2 claims, one spurious.
>Each firm knows that flooding the market would hurt all of their profits, so a form of unspoken coordination can occur, and this is next to impossible to prove.
See: Have been burnt by over production before. And Unspoken coordination is hardly an issue. Reacting to private information is bad, reacting to public information is normal and expected.
>It's hard not to see this supposedly coincidental aligned strategy of restraint and wonder if there's something more at play. All of these actions support pricing stability (for those companies) and suggests that no one is "breaking ranks" to grab a larger share by undercutting prices.
You are telling me that you are suspicious of an industry that has expanded, caused a glut that hurt their own businesses and killed their competitors, and are shying away from repeating that event that you outline as a clear mistake.
If theres so much money on the table, you will need another vendor. New RAM vendors have to weigh risk, including all the risk you have outlined, against what is likely a very massive capital outlay.
I don't understand when people blame AI for buying DDR5 DRAM - aren't they mostly interested in HBM? Or is the fab space being diverted to manufacture more HBM than DDR DRAM previously?
This quote is basically the TL;DR:
> memory suppliers have both the motive and precedent to coordinate behavior, even tacitly, in order to keep prices high. When only a handful of firms control the taps, it doesn't take a formal cartel for them to collectively benefit from constrained supply. Each firm knows that flooding the market would hurt all of their profits, so a form of unspoken coordination can occur, and this is next to impossible to prove. The backdrop of past cartels makes it hard not to be cynical when hearing that "AI demand" is solely to blame for increased prices. Whether or not any collusion is happening now, it's clear that memory companies are profiting immensely from the current crisis. After bleeding financially during the last oversupply downturn, the major DRAM makers are now seeing record-high earnings in the third quarter of 2025 thanks to the price surge, and to put it bluntly, the shortage is great for business.
> Each firm knows
In some cases, firm == family, https://en.wikipedia.org/wiki/Chaebol
I'm so excited to read the pro-monopoly take on this.
Not pro monopoly as such, but in the recently posted article on UK shipbuilding they mentioned that a major reason for it's decline was it didn't build out capital capacity after being repeatedly burned by cyclical nature of the industry, with some quote even claiming they thought they would be best positioned for the next bust explicitly.
What choice do these companies have, even assuming the demand doesn't crash? They can build out and cause a glut in a few years. They can not build out and then what? Barriers to entry are high but not insurmountable; long term they'll go the way of UK shipbuilding. It might take a while, though. Also one defector building out would have massive advantage, so it doesn't seem like a stable equilibrium. See also constant bickering in the OPEC, an explicit cartel.
I work in the refurb division of an ewaste recycling company. I was tipped off to these insane prices about two months ago, and switched focus from laptops to drives and RAM, and I'm starting to make bank. Roughly a petabyte of hard drives showed up two weeks ago, and I'm almost halfway through preparing all the drives. Hopefully I'll be able to finish and sell them in time to have a very merry Christmas.
Companies will always seek to maximize revenue. Your options are 1) don’t buy 2) choose competition.
And the competition doesn't exist anymore.
Glad to see I'm not the only bringing up just how untrustworthy these memory companies are.
I keep hearing about how the supply and demand cycles are “lagged” and that this price spike happens every 3-5 years for purely economical reasons. I feel like I’m being gaslit — I don’t recall any such price spikes in the past of this magnitude. You’re telling me that there’s a totally predictable price cycle and NO ONE has prepared for it? Or else prices are just high temporarily and no one can step in to increase supply? Either way, it seems that there’s not enough competition in this market.
Okay, this will be a bit on a conspiracy theory side, but there was a paper recently describing how to do matrix computations on a RAM stick connected to FPGA, and they've shown it's possible to do it cheaper per flop that GPUs. Except, of course, there's a variety of RAM producers.
It might either be artificial to keep GPU prices where they are, or someone already started building their RAM-based AI datacenter.
I've seen a fair number of articles suggesting that the finances of AI companies have lost touch with reality and that the AI sector is now well within bubble territory.
If AI companies continue to scale up and buy massive amounts of memory as prices spike, how much will that intensify the spike? Could feedback of this nature cause a price shock sufficient to pop the AI bubble earlier than it might have otherwise? How soon might that happen?
Hey, at least a fab didn't have to burn down this time. That's progress.
I've maxed out the RAM on every (consumer mobo) build I've ever done, and have always ended up appreciating it. I did a Ryzen build with 192GB back in January, so despite the questionable signal integrity of DDR5 especially on AM5 (more than one DIMM on each channel means lower speed) this time turned out to be no exception. I also stocked up on 130TB of HDD and 30TB of SSD, as it was clear we were headed towards some kind of economic disaster. (look, I'm avoiding politics!)
But the best RAM purchase I ever did was during the dot com crash glut a few decades ago. I maxed out my Athlon XP with 3x 512MB sticks for $27 each. A gig and a half of RAM. Those were the days.
What do you do with 192GB RAM?
My main desktop has 256GB, and my "router" has 128GB. Mostly used for software builds (I use NixOS as a source distribution) and VMs (which themselves are mostly used for web browsing, surprise surprise). The desktop had been fine at 112GB because that was the max that coreboot could train, but the coreboot code has been updated and used DDR3 sticks are cheap. The goals of the new 192GB machine are more of that existing usage, plus LLM experimentation and possible VFX rendering/video work. It's speculative of course, but as I've said I've never regretted bulking up on RAM. If it ends up feeling too vacant and echoey, I can always grab another motherboard and processor and make two 96GB machines.
Is RAM the new oil, then?
I’m sad the mid to late 2010s alt reality era has come for HN the last 6 months, after 16 years here.
This was my last respite for intellectual argumentation.
Now, every day there’s multiple stories where you have to be caught up on this cinematic universe where AI is fake and doesn’t work and no one uses it and anyone who does is a grifter and or amateur and or embarrassing and any data centers they build will be a waste and they’re probably not even being built and OpenAI is JUST like pets.com so this is basically the web bubble from 1999…so therefore, $X! (In this case, X = RAM supply shortage is fake and actually just coordinated price gouging)
As my MD friend noted wisely a couple weeks ago: it’s noteworthy how this became a culture after LLMs became ubiquitous and user friendly. It was tons of fun and happy times when we were going to reduce # of radiologists, not software engineers.
US gov wouldn't let AI bubble pop, the amount of money in circulation would make the US set back 5+ years and potentially into great depression
the aftermath for tax payer and your 401k would be devastating
Not saying that you're wrong, but of course if there is a real AI bubble, then "not letting it pop" only escalates the damage when it does eventually pop. The best outcome all-around is for the market to form accurate expectations quite promptly about the real potential and capabilities of AI, regardless of the immediate consequences.
My own wild guess is that this spike in RAM and storage costs is more of a potential drag on the tech sector as a whole than AI companies specifically. Maybe we'll see some systems being reengineered to cut the waste and the technical debt throughout and be a bit leaner and meaner, since that will be making a real difference to the bottom line.
or there is third way, AI company would achieve "near AGI" and "only" eliminated 80% job market
If/when the AI bubble bursts, there will be a lot of cheap used parts available to buy.
Is it actually standard modules that are high on demand, or are the chips directly soldered to custom mainboards?
You’ll own nothing, not even your own short term memory, and you’ll be happy.
you will own liabilities and be happy
Why would you need short term memory when you'll have AI agents /s
"For context, these increases have even outpaced the surge in gold prices over the same period"
And you can't just manufacture or not the gold supply into scarcity
yes you can. It's called mining
actually the problem was population boom. FDR raised the rate from $20 to $35/ounce, and by timeNixon took the USA off standard the global population 3x. So gold is no longer scarce because it isn't backing real currency anymore. Its just metal
Maybe the crypto bros reallocated from bitcoin to DRAM.
[flagged]
Might we perhaps see some meaningful investment into long-term alternatives to DRAM as a result of this price spike? Will Intel bring back their Optane persistent memory? What about HP's memristors/ReRAM? Magnetic core memory/MRAM?
This is what I'm hopeful for, but haven't researched enough into whether there are hardware startups in this space as I believe the majors will chase the hype cycle. Optane being canned always felt like a poor, shortsighted decision to me.
Add to that the fact that you can now buy your way out of US antitrust enforcement by taking to the right guy.
[dead]
[dead]