Average DRAM price in USD over last 18 months
pcpartpicker.com272 points by zekrioca 9 hours ago
272 points by zekrioca 9 hours ago
Funny how many dot-com esq things are popping up.
The big news story this cycle is that OpenAI secured two large contracts with two separate DRAM companies at the same time, carefully timing the deals so that neither DRAM company knew about the deal with other. Had either company known about the true demand from OpenAI they would have charged a lot more, but each only saw about half.
In other words, there was no collusion between the DRAM manufacturers. They were both caught off guard and left a lot of money on the table.
The current price increase is the result of the huge demand spike. Production takes years to ramp up, but demand has spiked rapidly. Supply and demand.
One difference that strikes me with the .com bubble is that I don't remember the .com companies having sustained multi-billions losses / cash burn. They were not profitable but this is quite different. If (or when) the music stops, won't OpenAI go bust immediately? That's quite a counterparty risk those companies are taking.
OpenAI is ten years old, dotcom companies were 2-3 years old.
Some dotcom-boom companies that survived also had sustained multi billion dollar losses afair - Amazon and Uber for example.
To those who haven't heard how colossal the size of OpenAIs contracts are.
900,000 wafers monthly. Tom's hardware estimates that is equal to 40% of global dram production capacity.
https://www.tomshardware.com/pc-components/dram/openais-star...
At that point even if the AI bubble bursts you have a solid business as a RAM scalper
They are not going to repeat the Windows 8 debacle again are they? (Over producing in response to perceived demand only to not get the demand when they are ready to sell) Didn't that bankrupt some memory manufacturers?
And when demand eventually crashes, all the new production capacity is left without buyers to sell to, so maybe it does not even make sense to create it.
There won't be any new production capacity until DDR6 comes out. There is no point in investing into an obsolete technology like DDR5.
China will happily take that business. They are producing DRAM now arent they?
You have to wonder whether they realy intend to use the ram, or they just spotted an opportunity to corner an essential market and choke/extort in resale.
Do they have the extra money to play games like that? I thought they were pulling back on ads because of the community reaction even though they wouldn't be introducing it now unless they needed it.
To put this in long term relation.
Even current DDR4 3200 DIMM prices are at an all time high.
These are 6+ year old chip specs now!
I even thought stuff was overpriced four years ago in mid-2021 already, but this is a whole new level.
Some sample long term data for those:
Between inflation generally, and DDR4 being obsolete and unsupported by current desktop or server CPUs, it would be unsurprising for DDR4-3200 DIMMs to be at an all-time high even without the current DRAM price shock. You can never count on old memory types dropping to bargain prices, because the major manufacturers are always eager to migrate the bulk of their production capacity to current-generation memory.
Dang, I'm going to have to put my old RAM on ebay. I thought it was worthless, but I was clearly wrong.
Now happy I bought that 256GB of DDR4 with the MB/processor combo of ebay last summer.
anyone want to buy a 2x16 GB DDR4-3200 kit that only fails memtest86 some of the time?
I wonder if semi-reliable RAM could be made to work for training. After all gradient descent already works in a stochastic environment, so maybe the noise from a few flipped bits doesn't matter too much.
Maybe it's time we make a simple web page 100KB again? Is there some kind of CDN minification, adblocking and compression service? Maybe even server side rendering of websites?
Then a smartphone would work fine with 1GB of RAM and everyone could be happy.
Simple web pages can be made small with modern tooling, e.g. AstroJS.
The problem is that most web pages these days fundamentally are not simple.
Rather than trying to make web pages small, the real effort would be in designing web pages to be simple.
The large majority of software devs, PMs and the like don't really know how to do anything else than a Node + React webapp.
I've come to the opinion that for the vast majority of apps I've built, it could all be built using HTML + CSS (all built server side). I can sprinkle in little bits of interactivity using something like HTMX. And I'll have a website that is very easy to optimise, has phenomenal backwards compatibility, and gets rid of a whole class of issues associated with SPAs.
I often regret in my career not pushing back more on "requirements" that ended up requiring a more complicated app, whereas the customer would have been happier with a simpler solution.
I guess you're right, but it's more of a curve, though. Once you get to any decent level of complexity, it actually helps to have a framework instead of just going all HTML+CSS. Also it helps having something standard as react (that every web developer should fundamentally understand) than doing your custom stuff if other people will be working on it in the future.
There's a lot to say about the side effects of frameworks but there's a reason why everything converges towards that.
Most people have phones that can handle webpages with 1-5MB JS bundles. Why artificially limit what you can do on the web? Why limit ourselves to 1GB RAM when more resources means tech becomes more useful?
Returning to simple webpages is popular idea on HN but it’s like wanting a car with no backup camera and crank windows. If your goal is to have your car be as simple as possible, then sure, but that’s not the case for most people.
Most people want their cars to be safe and convenient, and their webpages useful and rich, more so than they want to return to some idealized simplicity.
A simple webpage or blog with minimal styling that runs as an ARM binary on a TV remote is cool and fun but it’s not economically useful. It’s the equivalent of a manual scooter. We can build better apps (in the same way that car manufacturers can build less crappy infotainment systems) but optimizing for scarcity isn’t the answer in a world where abundance tends to grow.
(Edit: your downvotes mean nothing to me, I’ve seen what gets upvoted!)
Actually a plain car would be great
I crank the window up and down 3x faster than the little button
And I could adjust my damn seat before electricity is available... sigh
I’m glad you want that! But most people wouldn’t. Also, electric seat adjustments give you way more options than the manual adjustments could.
Your mistake is assuming there is some correlation with usefulness and size.
The JS Gmail UI from 15 years ago was just as functional as the one today.
Websites that are supposed to be simple lists end up bloated and laggy because of really poor JS that makes one request per item iteratively to populate a list.
You can do a lot with little, it just requires investing more in development which understandably most companies are uninterested in. Besides, plenty of websites are bloated as all hell. Why does a newspaper website, for example, have to be very much more than plain html?
Newspaper websites are a good example of bloat, true. I think if you’re in the business of primarily serving text content and not doing much interactive stuff, you don’t need a heavy site. A lot of them tend to cram their websites with trackers and ads and I guess that’s a business thing.
Tbh, it’s unpopular around HN, but I felt like AMP was a great experience for users. AMP pages were super fast and had no annoying banners - and none of my pet peeve: layout shift.
And if someone could make a chat program not use 1 GB of RAM that would be great.
I just checked my go-to stores in Europe. Holy shit. I bought a 5200-speed 16GB DDR5 kit in October for 55 eur. Now it's selling for 240 eur. The 6000-speed model is now 390 eur.
July: 172,38 € for a 2x32GB DDR5-6000 kit
Best price for the same one now: 689,55 €
I've been meaning to get more components for a diy NAS since atleast the last year and just been pushing it lazily. I'm literally kicking myself now when I actually started looking up deals for this black friday.
How to get every tech enthusiast in the world to root for OpenAI to implode in order to release the RAM hostages.
Not only that - I think OpenAI and all those other AI blobs must pay compensation damage to The People for inflating the prices here and skyrocketing costs.
https://www.mooreslawisdead.com/post/sam-altman-s-dirty-dram...
On October 1st OpenAI signed two simultaneous deals with Samsung and SK Hynix for 40% of the worlds DRAM supply... the shock wasn’t that OpenAI made a big deal, no, it was that they made two massive deals this big, at the same time, with Samsung and SK Hynix simultaneously! In fact, according to our sources - both companies had no idea how big each other's deal was, nor how close to simultaneous they were. And this secrecy mattered. It mattered a lot.
Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different. It’s entirely conceivable they wouldn’t have both agreed to supply such a substantial part of global supply if they had known more...but at the end of the day - OpenAI did succeed in keeping the circles tight, locking down the NDAs, and leveraging the fact that these companies assumed the other wasn’t giving up this much wafer volume simultaneously…in order to make a surgical strike on the global RAM supply chain…and it's worked so far...
OpenAI isn’t even bothering to buy finished memory modules! No, their deals are unprecedentedly only for raw wafers — uncut, unfinished, and not even allocated to a specific DRAM standard yet. It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM! Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!The cynic in me thinks this would be a convenient way for these memory producers to manufacture demand, while also making OpenAI look good on paper. It’s not like they haven’t been caught price fixing in the past. Win win for these companies and a loss for everyone else.
If manufacturing fake demand (warehousing 900,000 memory wafers per month?) doubles and triples customer prices, it would decrease real demand, hurting the manufacturers after the artificial demand ends.
Fake demand? Either they’re selling the RAM or they’re not. They don’t make money by pretending to sell into fake demand. They make money by selling chips. A sale is a sale.
> Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different
Wouldn't this be ... collusion?
Implicitly arguing that the memory oligopoly should have been coordinating is ... quite something.
OpenAI may well be doing something anticompetitive by cornering supply to foreclose competitors, but saying "they tricked the suppliers into not colluding!" is certainly a take you can have I guess.
Every normal market interaction could be described by scary monopoly-sounding words. You heard about you're friend's salary in the same industry, and ask for a higher salary next time you negotiate. You're colluding with your friend to price gouge your employer!
In reality, so-called collusion is normal and unobjectionable. But when price surges happen, often due to factors outside of the seller's immediate control, people look for any reason to find an ethical dimension and find how to place blame, because this is more convenient. Things that were normal become abnormal and suspicious. It is in consumers economic self-interest to act in this way, because it often secures favors to them from various economic policies that they don't normally get when the market is "normal". This is no less a form of collusion than what sellers might do to secure their economic advantage. But a key difference for these anti "gouging" policies is that it gives consumers a special privilege and makes market pricing less able to fulfill its social functions.
No, it's not collusion to ask for more money from OpenAI if you hear that they are trying to buy 40% of the world's supply. Increased demand leads to higher prices, that's normal.
OpenAI, by doing simultaneous deals, hid the true demand from the suppliers, thus lowering their price and raising everyone else's.
A case could be made though that it is OpenAI who is operating illegally, attempting to corner the market.
There is nothing suspicious or abnormal about this behavior. It is called competition. Ironically, trying to prevent this kind of behavior prevents competitiob, and is a key factor for causing monopolization
It depends... if OpenAI bought the DRAM in order to use it, then fair play to them.
If they bought the DRAM in order to stop their competitors from using it because they are falling behind, that's anticompetitive in spirit, though I'm not sure if it actually breaks any laws.
Knowingly attempting to buy or sell in quantities likely to move markets, for direct profit, is called manipulation and is most definitely illegal. this is true in physical markets commodity markets and financial markets. Not saying that this is what openAI is doing but it definitely merits an investigation.
What?? Replace DRAM with peaches or table legs and you have the exact example they give you during management training to explain implicit collusion.
That's when you raise prices without a change in the market conditions.
OpenAI is creating more demand, therefore the price must go up, if it didn't then there'd be shortages.
They have, rather famously, been caught doing financial shenanigans before. [0]
[0]: https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
Only 50-60% of the recent price appreciation is after October 1st, so this is not likely the biggest direct cause, more likely it's everyone buying the finished stock on the market in a frenzy because they anticipate things like this will impact availability in the future, and it becomes a self-fulfilling prophecy.
Sounds like OpenAI might be trying their hand at TPUs, like what Google has. They are one of Google's biggest advantages in AI right now. It would also give them insurance against Nvidia being everybody's hardware supplier.
This actually points the opposite direction, to doubling down on commercial GPUs.
NVIDIA recently told their board partners that they will need to source their own RAM and will not be bundling it with chips anymore.
If there is a supply crunch on DRAM, commercial GPU production lines will start having idle downtime. That is literally the worst possible thing that can happen to a company that has invested heavily in tooling and they will negotiate at or below cost production runs to fill the gaps if a customer can bring their own DRAM to the table.
Novice question: If they built something other than classic dram modules with the wafers maybe they could achieve faster bus speeds? How does Apple do it?
shorter traces than soldered DIMM allow higher MT/s, this is fixed by CUDIMM/CAMM2, the other part of this is # of memory channels on the board, not sure why, but most consumer DDR5 boards have been 2 memory channels, you need to go to threadripper to get 4 or 8, It's unclear to me if this will still be an issue with future platforms.
You're not paying enough attention to the performance and cost impact of connectors and sockets. CAMM2/LPCAMM and CUDIMM have yet to be demonstrated operating at speeds that speeds that match the fastest soldered LPDDR, let alone GDDR; there's still a clear advantage for soldering memory.
CPU sockets with more than two memory channels are also far more expensive; the higher pin count usually increases the number of layers the motherboard needs, and the larger size of the socket requires more metal for stiffening (and EPYC CPUs still have issues with imperfect mounting leading to some IO lanes not working).
Using BGA soldering for both the processor and the memory sidesteps a bunch of engineering challenges.
I think most of us came to the conclusion that the AI hype is heavily responsible for this.
I want my money back. There should be an extra-tax on all those AI companies - they are heavily responsible for DRAM costing more now.
I don't understand this. I'm looking at prices on some Asian store fronts and it's nowhere even remotely near these. I'm looking at DDR5-6000 2x16 for about $130, with no apparent limits.
Even with tariffs, transport, and other fees, you could get this to the US for way less than $400. I doubt the market could be this inefficient - in other words I don't think I just found a get rich quick scheme. So, what gives?
The current prices are a response to future stock, not current. It's 100% retailers price gouging with their current stock that they got for cheap because they know there will be limited stock in the near future. Asian retailers may be more honest and keep their margins the same, but will catch up in a month or two.
It's just surge pricing to manage demand, not "price gouging". Being able to buy RAM at high prices is a lot better than being unable to source it at all because everyone else is panic-hoarding.
We don’t really know whether that’s true, since it’s hard to prove a negative (i.e., suppliers aren’t colluding). But given their history of price fixing it may be worth looking into.
> It's just surge pricing to manage demand, not "price gouging".
This carries the same energy as company leadership insisting that a RIF is not a layoff.
Price gouging has a specific meaning. It’s when costs are raised rapidly only because people are desperate and have no other choice. (e.g. water after a disaster)
Retailers raising prices in reaction to an incoming event that will take supply away from everyone is not that.
Is it better? As with all price gouging, better for those who can afford it, sure, but not for those who can't. The proper way to combat scalping is to implement fair allocation methods (for a start purchase quantity limits) and punish people for scalping.
Look at how most places handled war-time gasoline shortages. Rationing coupons, purchase limits, demand leveling (like the odd-even system), price or profit controls, strict prosecution of scalpers and price gougers. And it's not like only the communists did this - even the US had most if not all of these things. And it worked far better than the shit that happened during the pandemic shortages. Governments used to know how to govern.
The high prices ("price gouging") perform a social function because some cannot afford it; they prioritize what matters to them more, and ram is left available to those who absolutely require it. Trying to get around that by forcing prices below the market price simple encourages scalping behaviors. If prices are below the market price, but at the market price you would prefer the cash in your pocket over the ram stick, then you have every reason to sell it higher than you bought it because people are willing to buy it. It is those willing to buy it that are the main culprit in establishing the true price.
In reality, it is almost never a true binary of "afford" or "cannot afford" like critics of surge pricing make it out to be; people evaluate the price according to their circumstance and make a trade off. It is because of these decisions, the state of demand, that surge pricing is possible, not because of the machinations of evil price scalpers. That is why manufacturers couldn't lower prices even if they wanted to; gpu msrp being a great example of gpu vendors being caught between consumer ignorance about economics and the facts of reality that gpus are scarce enough to warrant higher prices.
I'm definitely not saying it's a binary can/can't afford situation. The point is that people don't "afford" things equally. I might some need RAM so badly that I'm prepared to take a huge risk and spend half of my paycheck on it. But that doesn't matter because someone else has 3x my paycheck, savings, and investment portfolio and a good credit score. How much money someone can spend on somethong is no indicator of how much they need/want it.
Something like "GPUs are actually scarce" doesn't even make sense to say, since scarcity is more a function of demand than supply. The supply of GPUs wasn't exhausted because people suddenly needed more GPUs or because Taiwan couldn't produce as many of them as they used to, it was because a few rich bastards were buying into a bubble so they could make as much money a possible before it all comes crashing down. They didn't "need" those GPUs much more than even the scalpers. They were just a vehicle to make short-term profits at the expense of everyone else.
And yes, of course those willing to buy things are the ones enabling the peice gouging. But that's not a useful observation. You either need something, so you'll buy it even if it doesn't make financial sense, or it makes financial sense to buy it, so you will. Notice how scalpers also fall into that second category, along with the rich bastards draining the supply.
If you want to be "fair" for a necessity such as gasoline, you can have tradable rationing coupons. That way you are rewarded if you buy and use less gasoline, but the excess windfall due to the shortage is still transfered to you and away from the supplier. But even this assumes that gasoline is in fixed supply and there is no way of increasing its total production by paying more, which is not a very good assumption.
In a time of shortage, throwing more money at the problem usually won't increase supply. A shortage necessarily means that if you make more of something, you're guaranteed to sell it basically instantly, so there's already an incentive to increase production.
And it's not like the higher prices mean more money goes to the producers so they can invest in more production capacity. The price increase is spread out between every middleman in the chain untils there's almost nothing left. This could work only if the producers themselves are the ones raising prices, but then everyone else would still add their own cut, leading to even crazier price hikes, and also it's unlikely that extra profit would go to much more than lining the owners' pockets.
Additionally, demand spikes usually don't last, so any new production capacity you build will be a liability later, after the market settles down.
It's speculative price gouging. Calling it "surge pricing" doesn't stop erosion of consumer trust in the market. Watch now as more people more readily jump to price fixing conclusions. Not helped by the inevitable further increase in speculation through feedback loops and the resultant volatility.
First come first served is a better principle than "surge pricing". A lottery is a better principle than "surge pricing". In the case that someone over purchased, they're free to dispose via secondary market if the value to them is lower than the out of stock price. I.e. decentralised pricing (and profits). Secondary market sales are just more efficient, they occur at negotiated prices that reflect true individual valuation, not the retailer's speculation.
I'd rather reward diligence and personal responsibility - if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price. Rather than passive reliance on wealth to solve problems. First come first served values effort and foresight. Scarcity is managed through time and effort rather than money.
> First come first served is a better principle than "surge pricing".
This is called a price ceiling, and it's a bad idea with a track record of failure and significant harm.
I'd rather pay extra and get what I need with 100% chance than get what I need cheaply with 5% chance and otherwise be forced to go without or buy from scalpers for the same price I would have paid anyways. This is the purpose of prices. So the people who really need it can buy it, and those who are borderline about the purchase decide to opt out.
If you're concerned with wealth inequality or one large buyer cornering the market, there are better ways to address those problem than prices ceilings.
> This is called a price ceiling
The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
> I'd rather pay extra and get what I need with 100% chance
False dichotomy. Neither approach increases supply. Of course according to economists who can hand wave away bullwhip effects with simple "this model assumes X" statements that go unquestioned in the conversations which cite the findings of the given model but i digress. According to economists, both approaches do increase supply, the theory goes that the price gouging retailer invests in more factory capacity. Or the factory owner buoyed by vibrant secondary market activity views increased production investment as a safe bet. Maybe there's some truth in the latter...
> If you're concerned with wealth inequality
I'm concerned with lazy financial engineering over hard work. Why should the scrappy but innovative startup be excluded from resources over the sclerotic incumbent with a deeper wallet?
The bullwhip effect is the whole reason why retailers may want to hold greater product stock in the first place; to absorb transient demand fluctuations and not have to pass them on in full to the supplier.
And a "scrappy but innovative" startup has an obvious interest in being able to source the DRAM or other goods they require, even at higher prices.
> The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
What is it then? If you allow the price to increase, there is no need to enforce a different rationing mechanism. The only reason to think of implementing the alternate rationing mechanism is if the good isn't being rationed. If you insist on the previous, lower price, then that is a price ceiling in so many words, with the consequence of needing your rationing mechanism. If you allow a higher price, then your alternate mechanism is unnecessary.
> The act of eliminating surge pricing is not a price ceiling.
If you force companies to price products lower than what they want, then you have a price ceiling by definition.
> Neither approach increases supply.
I didn't mention or allude to supply at all, although it's true that price ceilings also decrease supply (less so if there is monopoly or collusion, but they still do).
I was talking about resource allocation. The person who needs a new system because their previous computer broke will be willing to pay the extra money, but the person who already has a DDR4 system with a 5950x that runs their games well enough will be content to hold off on their AM5 upgrade to DDR5 because the marginal improvement isn't worth the extra $400.
If you have a price ceiling like what you proposed, the person with the DDR4 system may buy the DDR5 that they don't really need 1 day earlier than the person who actually needs the DDR5, creating a misallocation of resources.
(that's an example of the more abstract principle at play).
Theoretical arguments aren't even needed here. The empirical history of price ceilings is there for you to google.
If you didn't know that what you were proposing is a price ceiling, and you thought that I was talking about supply instead of resource allocation, then I mean this with no offence intended but you should study elementary economics before forming confident opinions on the subject. Society is in a vulnerable point with cost of living pressure and we don't need more energy behind these harmful populist ideas.
> It's speculative price gouging ... if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price
The word for "monitoring market trends, anticipating needs and acting quickly" is, well, speculation. Why should a retailer not be allowed to speculate and hold more product stock when they anticipate a future crunch? In fact, the whole reason prices have become so volatile right now is that this supply crunch was not properly anticipated.
The reason why retailers are not "price fixing" is that price fixing involves setting an artificial ceiling on total production; retailers are not in a position to do this, and there is no evidence at all that DRAM makers are doing this either.
> First come first served is a better principle than "surge pricing"
First come first serve must means enterprising individuals would buy as much as they could afford and then immediately relist it on Facebook Marketplace for the actual market rate.
Thinking that retailers could keep RAM prices low and also keep it in stock for you is irrational.
>> Thinking that retailers could keep RAM prices low and also keep it in stock for you is irrational.
Not a claim i made
> It's 100% retailers price gouging with their current stock that they got for cheap
The retail price of a product is a function of the market rate at the intersection of supply and demand. The price paid for inventory on the shelf doesn’t matter.
It works both ways. If retailers bought a lot of RAM at high prices and then the market suddenly dropped, they could have to sell it at a loss.
Some people get irrationally angry at this, but you do it too. If you bought a house for $500K and the market went up such that it was worth $700K, you wouldn’t think it was “price gouging” to list it at market rate. You’re just trading an asset for cash at the current price. The price you bought it at is irrelevant to the price you’re going to sell it for.
Aus stock was cheap a couple of weeks ago, it's caught up though. I'd snatch it up if you need it.
The market is inefficient. Also, if you are in a small country, the demand is not there. So the seller will sell with regular prices. The stock will also be very low to export it back to where it came from.
I upgraded my pc about this time last year (new CPU, RAM, and motherboard).
I was originally going to just get 64GB of DDR5-6000, with the option of adding another 64GB later, thinking the price might drop even further. At the last minute, I decided to get the whole 128GB instead. Glad I did.
If these prices don't return back to normal I just don't see how Valves steam machine is less than $1,000 USD.
When Gaben's 11 organize a heist to steal the RAM wafers from the high-security vault in Sam Altman's basement, the price of the steam machine will drop below $1000 USD.
They would already have contracts in place.
As someone currently dealing with pricing changes for DDR5 volume contracts (admittedly, only ~30k DIMMs, but still), those contracts are a lot less rigidly priced than we (engineering) realized at the time - “agreed price” is a lot more... flexible... until the pallet is delivered. Especially because our contract is with our manufacturer who has their own contract with a DIMM manufacturer who has their own contract with the DRAM manufacturer. DRAM is substantially more like a spot market than the average consumer would expect (at least at my volumes). The same is true for my HDDs (~100k unit/year contract) and my CPUs (10k unit/year contract). At least for HDDs our contract quantities are getting honoured (and we actually still have Seagate and WD fighting for our business), which I’ve heard hasn’t been the case for smaller orders.
Yep. It's a bit like some gas projects when prices spiked due to Russia's attack on Ukraine.
Companies had supply agreements in place but they will find essentially any excuse to "delay" delivery. You might eventually get your product but fat lot of good it does for you while you're mired in court and not getting anything anyway.
The particularly amusing example is a project continuously selling spot LNG while saying it's not yet operational, stiffing the companies with which they'd signed long term contracts with on a technicality.
For a while, yes. But I'm sure they will want to sell it for longer than a year or so.
Even Samsung is running into this issue now: their own internal foundry is refusing to give them a long term contract now so the S26 series will become more expensive.
If this happens to Samsung, what leverage will a player like Valve have?
> Even Samsung is running into this issue now [...]
These large corpos are so greedy to the point they harm themselves. I remember something similar with Amazon, where the Amazon shop had to redo the whole architecture from some microservice setup back to a monolithic approach because they used AWS and paid these weirdly structured prices like everyone else. Which made running a monolithic architecture under such constraints inherently cheaper. Not sure what the resulting compounded business cost for this "endeavor" was, but more often that not such things are never accounted for, so they don't show up as an issue.
That's their costs, not our costs. Regular PCs will cost more, which means Valve can price their consoles higher and collect margin, because there's no other competition/substitutes.
Maybe, but from a platform and brand development perspective, there may be smarter strategies.
Their interest isn't necessarily in squeezing out the most margin on each unit. If unexpected market conditions let them more or less hit their original margin targets but get far more units installed in homes, that could be much better for them in the long term.
There's a lot we can't know as outsiders, at the moment.
> that could be much better for them in the long term.
i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Valve's existential threat is from microsoft closing off windows somehow (or extracting rent...like a store!). Therefore, pushing steam machines which can be run without windows, is both business expansion as well as an insurance policy.
> i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Wasn't that true with the previous incarnation of the steam machine, the valve index, steam controller, etc.? IIRC their VR gear was some of the most expensive consumer level gear on the market.
As I mentioned in a prior post 7/18/2023 - G.SKILL Trident Z5 RGB Series 64GB (2 x 32GB) 288-Pin PC RAM DDR5 6000 - $253.
Today - G.SKILL Ripjaws S5 Series 64GB (2 x 32GB) 288-Pin PC RAM DDR5 5600 (PC5 44800) Desktop Memory Model F5-5600J3636D32GX2-RS5W - $620.
Prices from Newegg.
GPU prices never went down after the crypto rush replaced GPUs with ASICs. I hope this does not repeat with DRAM and AI.
A possible answer as to why this is happening. https://thememoryguy.com/some-clarity-on-2025s-ddr4-price-su...
That's from June 2025. It's not relevant to current situation — prices have been rising since October 2025 due to agreements that became public in the same month.
I wonder who the "visionaries" were that set the marked increase trend of max prices near 2024-12-09 for DDR4-3600 2x32GB and DDR4-3600 2x16GB?
I think that acted as a dog whistle for the rest of the market signalling there'd be no consequences for ripping other people off.
Also relevant: https://youtu.be/B7sB1-8jKno
Really hoping this comes down again soonish. My desktop doesn’t have many years left :/
Apple system config upgrades not looking so bad anymore
On this graph, the most expensive RAM is $12.5 / GB ($400 for DDR5-6000 2x16GB). Apple charges
- $25 / GB ($200 for 8 GB for the M5 MacBook Pro and the M4 MacBook Air)
- $16 / GB ($400 for 24 GB for the cheapest M4 Pro MacBook Pro)
- $12.5 / GB ($200 for 16 GB and then $800 for 64 GB more for the most expensive M4 Pro and M4 Max MacBook Pros)
and Apple's RAM is faster than PC RAM.
Came here looking for this comment.
I wonder how much price increase it takes for Apple to raise theirs.
Guess that’s the flip side of selling sometimes three to four years old hardware with zero price cut (I know they’ve been doing better in the M-series era)—the price also stays there when it’s skyrocketing elsewhere.
What specifically happened in June to set this off? It can't just be "waves hands AI/LLMs" because ChatGPT has existed for years.
> What specifically happened in June to set this off?
Tariffs implemented by this administration:
"Inflation has begun to show the first signs of tariff
pass-through," said Ellen Zentner, chief economic
strategist at Morgan Stanley Wealth Management. "While
services inflation continues to moderate, the acceleration
in tariff-exposed goods in June is likely the first of
greater price pressures to come. The Fed will want to hold
steady as it awaits more data."[0]
Why did European prices have the same increase then?
> Why did European prices have the same increase then?
Where do Europeans get their DRAM from?
If it is the same handful of companies the US gets their DRAM from, then why would Europeans pay any less? Because the EU is not engaging in the same asinine trade war?
Sounds good in theory, but in practice those same few companies can set prices for markets outside the US to be at/near US prices. It doesn't take much effort for manufacturers to set their prices at or near those of their competitors and rely on an implicit mutually assured destruction[0] understanding.
0 - https://en.wikipedia.org/wiki/Mutual_assured_destruction
I'm sure if they didn't keep the prices somewhat similar, you would have a bunch of people in Europe selling RAM to Americans.
> I'm sure if they didn't keep the prices somewhat similar, you would have a bunch of people in Europe selling RAM to Americans.
I was just about to edit my response to the GP to say the same thing. Let's explore this hypothetical situation a bit further.
Suppose there was a DRAM manufacturer named "Acme DRAM" which decided to have a separate pricing schedule for the EU reflecting the lack of insane US tariffs.
Some enterprising entrepreneur in the EU would establish a company in the country having the least US tariffs and resell Acme DRAM to US companies. Surely this would make money hand-over-fist.
Problem is, the US DoJ does not look kindly on this kind of enterprise:
DOJ also has demonstrated a growing willingness to pursue
criminal charges against companies and individuals involved
in customs fraud schemes such as the purposeful
misclassification of goods, falsifying country-of-origin
declarations, and intentionally shipping goods through
low-tariff countries. Importers of goods into the U.S.
should expect criminal enforcement to accelerate in the
coming months and years.[0]
All of this risk in the pursuit of lower profit margins by definition.
0 - https://natlawreview.com/article/what-every-multinational-co...
Huh? It's not the manufacturers paying the tariffs, its the importer? US tariffs do not affect the margins of the manufacturer.
> Huh? It's not the manufacturers paying the tariffs, its the importer? US tariffs do not affect the margins of the manufacturer.
This is a case of second-order effects[0].
See this post[1] for details.
0 - https://research.gatech.edu/blind-spot-big-decisions-why-sec...
Afraid not. This has not happened across the board to all components.
> Afraid not. This has not happened across the board to all components.
Afraid so. Industry impact from tariffs does not require them to be applied "across the board to all components." See here[0] for more information.
With erratic massive tariff proclamations, counter-tariffs are to be expected. All it takes for companies to inflate prices is to either:
A) be provably impacted by tariffs
B) be opportunistic by being "tariff adjacent"
0 - https://www.tradecomplianceresourcehub.com/2025/12/03/trump-...
OpenAI bought up a big chunk of global DRAM production:
https://www.tomshardware.com/pc-components/dram/openais-star...
"Waves hands at AI/LLMs"
The problem is that training these models and using these models has required exponentially increasing amounts of memory.
ChatGPT has existed for years and in those years it's userbase and model size has increased tremendously. Not to mention the fact that a lot of competitors have sprung up in the wake of GPT. Including the likes of cloud based open model hosting services.
Been planning to build my daughter a gaming PC for Xmas. Feel lucky I found a pre-built with 32GB RAM in a warehouse store on black Friday that hadn't been marked up yet. First pre-built desktop I've bought in decades and it was actually cheaper than I could build myself due to the crazy RAM prices.
Time for Intel to re-enter the memory market?
It hasn't been long since this happened The price hasn't been rising all year For d4 it's June, d5 is September so what happens
So, RAM outperforms stocks, gold, bitcoin, NFTs, real estate?
I asked ChatGPT if it can generate a bunch of RAM for me. He says this is some kind of famous joke or something.
(I’m sorry I couldn’t resist)
"Downloading RAM" is more feasible than people think.
The seek time of a consumer-grade hard disk is said to be on the order of 10 ms. That's roughly the latency of a very high quality FTTH connection. Meaning that if you run a HDD rather than an SSD, a swap file in the cloud could potentially be faster than a local one (especially when you consider multiple reads/writes that could be done in parallel).
It's not exactly downloading more RAM but decently close to call it that for the joke.
Maybe this is a "strategic inflection point" for Intel... to get back into the DRAM business?
They should get back into the Optane business first. Take the existing Optane memory to a modern PCIe 5 bus, and it will be competitive with lots of uses for DRAM. PCIe 5 gives you basically DDR4-like transfer bandwidth, and the Optane media has better latency and far better DWPD endurance than existing NAND SSD's. It's a total no-brainer.
Why is it that consumers always get shafted here and we just keep going on like "it's just business"
Why isn't some consumer protection regulator going "actually no, OpenAI, you can't corner the market for the foreseeable future"
Hell, make OpenAI pay for this shit up front. You want to corner the market? Put up the money you don't have.
Who are the main players that make or benefit from these inflated DRAM prices?
the other side of the AI bubble are a huge amount of 2nd hand parts of all kinds going to come onto the market?
Yeah though expecting that to A) take years still and B) the gear will be old AF
Bit like you could get NVIDIA Server cards before things went crazy but they’re on ancient cuda etc so not exactly as glorious as one would imagine
Waiting for all my out of use 4, 8, and 16GB DDR3 DIMMs that I have somewhere in a drawer to become the new gold standard first.
I'm finding that getting an LGA socket to give me two reliable memory channels is far more valuable than the sticks that go into them. I've got at least two motherboards in use at my home right now with only one working channel of memory. There are sticks just sitting around.
Memory chips are divertied to registered server RAM and those will not fit most workstations
Nothing stops you building an EPYC based workstation. For the last socket you could even get mATX boards from Asrock Rack which were kind of ridiculous - I was very tempted to build my PC around one.
Unless it drags on for years so that the parts are old by the time everything gets liquidated.
A lot of the goods won't be essily repurposible for consumer or small business work loads.
Imagine if auto manufacturers all refitted their factories and supply chains to produce military vehicles for a war effort. New family cars would run dry, and when the war ended, some folks would figure out make clever use of some surplus military vehicles for street travel and commerce, but most of the surplus would just be shifted to other military markets and family car production would take some time to resume.
AliExpress has been selling 2010 or whatever Intel xeons in dual socket desktop board kits for gaming. they are fairly affordable and hold up to almost modern games so people that can't afford new gen systems.
You can buy old rackmount servers on ebay for relatively cheap and used them as desktop PCs AFAIK (though they can be on the noisier side).
All in a pretty bad state of wear, I imagine.
At least with ECC memory it's very obvious when its failing as you'll see reported correctable errors.
I would snag up every possible H100 and H200 if the AI bubble burst and their prices went into steep decline.
No, they'll be shipped off to developing nations to be dissolved for rare earths for the next boom cycle.
Memory chips don't contain rare earths, and e-waste recyclers pay a tiny fraction of what used-hardware buyers do.
Got real lucky. I was trying to negotiate a guy down on RAM. I wanted a server without as much RAM. In the end, I had to take the 768 GB. It only cost $1k more than 192 GB at the time. Unfortunately I can’t really sell off the excess because I need to substitute down the RAM modules to fill all 12. Ah well.
Stagflation!
I hate this. PC gaming is my hobby, the only one that’s lasted my whole life. It’s always been there. It’s how I met my wife. It’s how I relax after a long day. It’s how I’ve participated in so many stories that stick with me and given me so many memories.
All of it is being murdered by the AI bros. Before them it was the crypto bros. It’s one thing after the other and I hate it so much.
Just live with the PC you have for two more years. It's probably not a big deal? A moderately capable machine from 5 years ago is still marginally capable.
Hopefully this blows over in a few years and you’re right and I’m just catastrophizing.
One way or another. Either the AI stuff cools off, or the RAM people spin up more fabs.
The silver lining could be that game developers spend more time on actual gameplay, and less time on chasing graphical fidelity.
Silksong is playable on an 8 year old Nintendo Switch.
Yeah. Thanks for trying to keep me positive.
Hopefully this all calms down eventually. But it's hard not to feel like shit in this situation.
I mean its pretty rare to buy more RAM after completing your PC build + that single PC is going to last you 5+ years. Also mobos usually only have 4 slots in total so its not like its even going to take a lot. I'm rocking 2x48gb sticks and that's plenty for gaming.
The prices are wild tho.
I bought that ram in March 2024 for $384.81. Now it's priced at $1,172.99. LOL
It’s frustrating that it’s been one thing after the other, seemingly aimed directly at the thing I enjoy most.
There is more than a lifetime of incredibly great PC games that run on your existing hardware, and if this is your life's hobby, then paying an extra $100 or so every year or few is a drop in the bucket of your gaming expenses
I can afford it. Other people cannot, and the hobby is driven by a market existing for games. If newer people don’t enter the hobby as others die out, it fades away.
Increased demand for computer components for purposes other than gaming constitutes "AI bros murdering your lifelong hobby"?
PC gaming is not "murdered", it's doing better than ever.
In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
The inflation-adjusted price per gigabyte of RAM has dropped from $3/GB to $2/GB over the last 10 years, even including the recent price hikes.
So spare me the hysterics, your hobby is fine.
And you know what? The increased demand for compute always spurs innovation, so you'll probably get a better computer in the end as a result. You're welcome.
> In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
This is like saying "Spotify's subscriber count grew by 800% over the last 10 years. Music is doing better than ever!"
Inflation metrics seem to fail to capture the increase in leading-edge tech products of the past 18 months.
If we measure the value of the dollar against megabytes of RAM, well, in 01987, the difference between a StarBoard with "2 megs space" and a StarBoard with "2 megs installed" was US$484 (https://archive.org/details/amazing-computing-magazine-1987-...), so a "meg" of RAM cost US$242. Today https://pcpartpicker.com/trends/price/memory/ says "DDR4-3200 2x8GB" costs US$108; if that means 16GiB, as in 16384 "megs", a "meg" of RAM costs US$0.0066 today.
So the dollar's value has increased by roughly a factor of 36700 over those 38 years, averaging 32% per year.
That would be an average yearly inflation of -24%.
Too bad you can't live in DRAM or eat it when you're hungry.
> If we measure the value of the dollar against megabytes of RAM
We should be measuring it by the amount of RAM in a typical household PC in 1987 and today.
Even though a "meg" of RAM costs less than 1 cent today, I can't do anything useful with it.
Even if we are generous and buy a whole $1 of RAM today, it only gets us 150 MB of RAM, which, while infinitely more useful than 1MB, is still completely useless for running a modern OS/Browser.
What does your math say about that?
Math says that's stupid. You can still do everything with a "meg" of RAM that you could do in 01987. In fact, the "meg" of RAM you get now is enormously faster, and you can interface it to other things that you couldn't get at the time, such as SSDs, so you can do strictly more with it.
Economics says you're spending newly abundant resources freely in order to conserve those that are still scarce. Economics also predicts that people will adapt to RAM prices doubling by using RAM more frugally, spending more of other resources to compensate.
You can still run software from 1987 on a megabyte of ram, it just does different things from a modern browser.
Sure how does it compare to similar low, middle, high end cpu prices? NVME drives?
They seem to fail to capture a whole lot of things. Supposedly $1 in 2000 is worth $1.88 in 2025. So 88% inflation over the 25 years. Meanwhile the median home price has increased by 150%. Family insurance by 350%. Median college tuition by 225%. Childcare costs have risen by 200%. But sure. We can buy super cheap 65" tvs now. Hurray for us! Literal kings who lived hundreds of years ago couldn't possibly imagine a world with cheap large screen tvs. So the poorest among us should rejoice at the wonders they are able to enjoy while they skip meals and ration their insulin.
This explores the ideas behind your post: important things, like education and healthcare, have disproportionately risen in price while not-so-imortant thing have gotten less expensive.
https://www.yesigiveafig.com/p/part-1-my-life-is-a-lie
I don't exactly agree with the numbers, but I think the basic ideas are true
The inflation basket only represent a hypothetical average person, who doesn't even exist.
It's more useful to construct multiple separate inflation measures that represent different types of people. Like a "typical renter" inflation figure vs a "typical homeowner" inflation figure. It wouldn't be hard to do and would shine a light on inequality and help explain the rise of populism in certain segments of society.
An even better measure would somehow appropriately normalize the figure by the average disposable income in each of the segments to come up with a figure that measures the felt impact better.
The figure would be negative for wealthy people (who actually benefit from inflation because of asset price inflation) and positive for poor people (whose disposable income mostly goes to rent).
A dozen eggs is up over 350%, but a 6-pack of Budweiser is only up 60ish% since 2000. So you know, it all balances out. Maybe drink a few extra cans of Bud with your next meal.
[flagged]
What a ridiculous statement. We have a massive number of people in the USA relying on financial support for purchasing food. Not every single American is fat, especially those who literally can’t afford to eat.
There are lots of Americans who go hungry, but mostly outside the USA. Starvation deaths inside the USA are almost entirely secondary to medical conditions that interfere with eating.
the reality I observe does not align with your Dickensonian version of it. food is laughably cheap in the US.