If so, on a liquidation event, I'd imagine they'd have to be sold to compensate the shareholders.
Which option is a profit-oriented, not-totally-profitable AI outfit going to choose?
It's very simple: some is better than none
Clothing companies do typically put items on sale first. They'll even sell "at a loss"
That's why used enterprise/industrial electronics are way less common on EU eBay compared to other places, despite the industrial base being so large (or used to being so large). You can actually see when WEEE came into force because very little used T&M equipment entered circulation in the EU after the early 2000s.
Another factor is lease agreements, where companies don't buy equipment and the lessor "recycles" (i.e. crushes) the equipment to avoid creating a used/second hand market as an additional benefit.
https://www.tomshardware.com/tech-industry/data-centers-in-n...
It’s not the same as dark fiber capacity being overbuilt for demand that wasn’t there yet.
Is this really true? Other than the occasional outages, the frontier labs don't seem to be getting overloaded by API pricing customers at peak times, and with Opus 5.5 using even less tokens (which translates into less GPU time and less money spent on the same tasks), it seems actual capacity isn't literally strained right now. If current needs are met by currently available capacity, where is this demand going to come from?
Maybe somebody will use them, but not the original owners.
I can directly confirm this fact, and I'm just some random dude out here in suburbia, who can only imagine what the Indiana Jones warehouse of glowing GPUs looks like in a data center.
We're all salivating over this imagined future where an AI toilet seat greets us every morning.
Who here has already asked AI how to make your own DRAM ? I know you have. Thats a sign of peak interest in a subject or market.
I'm not, but I can certainly believe the surveillance industry vectoralists are salivating over a future where their AI toilet seats greet each of their data subjects every morning.
Today, the "old" corporations like google, meta, amazon are also spending a lot of money.
Not according to this article[1], which says:
"I believe we are now in an inevitable overbuild situation, one with no neat, tidy Dot-Com Bubble-style exit story. Demand for NVIDIA GPUs — and those from Broadcom, AMD and other semiconductor companies — is driven by speculative capital believing that the AI industry will become magnitudes larger than it is today, largely driven by the fact that everybody believes there’s far more demand for compute capacity than actually exists. ... NVIDIA has created a remarkable illusion perpetuated by the media — that GPU sales are a direct measurement of the actual demand for AI compute, rather than a measurement of how a few companies are willing to invest in an idea two years in advance, using circular financing as a means of creating the sense that you must buy these GPUs now, or you’ll miss out on the future. ... At the very least, hyperscalers are going to be burdened with brutal depreciation charges or onerous write-offs for years to come, whether their capacity turns into revenue or not. ... I don’t see how 90%+ of NVIDIA’s sales ever end up generating a single dollar of revenue, and considering the amount of project financing-backed data center debt deals, there’s very little that exists to protect investors if AI compute demand never arrives."
The man saw the market demand for saying "AI bubble is going to pop and AI tech is going to wither away and die", and went to meet it on the supply side - truth be damned.
So far, AI companies still keep getting bottlenecked on compute, AI utilization increases - driven by, among other things, increased price-performance of AI making it viable in more and more roles. And the demand for both AI inference and AI hardware in general shows no signs of stopping.
HN discussion: https://news.ycombinator.com/item?id=49526069
The other hyperscalers are even further delayed. CoreWeave has some VeraRubin, but not in volume (as I understand).
After the crash no one will need to build a new data center for 20 years, and there will be plentiful power connections
A lot of its current use is just basic questions that could be handled by local llms just as well, yet they are being processed on GB200s on the other side of the world.
It’s what everyone does, from company to individual. When you sell your used car or list your house, you price it according to what the market will pay. When companies sell a product, they list it for what the market will pay.
The conspiracy theories and accusations of collusion were flying everywhere until CXMT came online. It felt like every comment section and headline was predicting that CXMT’s arrival was going to crush the cartel and smash the “price fixing” when a new competitor arrived.
Then CXMT memory arrived at nearly the same price as everything else because, to the surprise of nobody with basic economic understanding, they were selling their products at market rate too.
The high price is being driven by demand combined with a supply chain that takes years to increase output. Just prior to this demand explosion, RAM and SSDs were so cheap that producers couldn’t justify building out more capacity. It would have been a bad choice at the time unless they had a crystal ball that could have predicted the future.
I’m willing to spend a lot to drive down the price. As a show of good faith, I promise not to actually pay any more than 50% of what you can get on the used market.
Obviously, spending additional dollars on purchasing RAM will drive the price of RAM up.
Just as obviously, spending additional dollars on manufacturing RAM will drive the price of RAM down.
This particular market is pretty clearly not functioning well, and there are textbook Econ 101 reasons why a market may not function well. Of those textbook reasons, I suspect the most obviously applicable one here is large barriers to entry.
Econ 101 would be understanding that you can’t snap your fingers and have a new DRAM factory online overnight, and that this isn’t an indicator that the market is not functioning.
You’re defining a market “functioning well” as meaning the price is low for you and all demand is satisfied immediately without raising prices.
That’s not really what is meant by functioning markets. A functioning market does allow the price to increase as this is what spurs investment into new capacity and new developments.
It also allows the DRAM to go to the most profitable activities. This is actually what the whole Georgism and Land Value Tax people advocate for with property taxes, which bizarrely many people think they like: The idea is to raise the price of something (in that case, your property tax) to ensure it’s not being “wasted” on an economic activity that isn’t optimal. The most optimal economic activity is the one that can afford to pay the most, which is the one that captures the resource.
It’s not fun when your use case isn’t the one that can pay the most, but that’s literally how markets work.
Wasn't there this whole legal thing a couple of years ago involving most of the dram manufacturers?
You can quibble over my usage of “functioning,” but all I’m referring to are the textbook characteristics of perfect competition. I’m not claiming that someone is to blame if any of these conditions don’t hold. Of course they never hold perfectly in practice!
I’m not trying to blame anyone for the fact that it takes a lot of money and time to get a new DRAM factory online. I’m just saying that this fact is bad for essentially everyone except for the small number of current DRAM producers.
> You’re defining a market “functioning well” as meaning the price is low for you and all demand is satisfied immediately without raising prices.
Well, yes, that’s pretty much the only reason to advocate for competitive markets, isn’t it? If you don’t want allocative efficiency and productive efficiency, then what do you want?
A calculated move. If they're right and the AI demand is a fluke, they don't overinvest in expensive manufacturing infra, they don't get saddled with more capacity than they need, and there is no memory glut eating their margins in the future. If they're wrong and the AI demand isn't a fluke, they get to pad their margins big time, and the supply bottleneck isn't their problem. As long as they shake hands and bet on "fluke" in sync, the downside is very limited for all of them.
They shook hands, they bet in sync, and they were wrong. Now the supply bottleneck is everyone's problem.
Potentially even theirs - if Chinese vendors can use it to wedge themselves into the market when it's at its most lucrative, and invest in growth. Which they almost certainly will.
Let's not rewrite history: Sam Altman kicked this off by trying to corner the memory market to get an edge over his competition...in return for only a year's supply of memory. The resulting supply squeeze has been very profitable to the RAM manufactures, and will last many multiples of OpenAI's initial contracts, so they can't be too mad at him for bot disclosing his other wafer deals during negotiations. They can afford to gift him a thank-you yacht named the "Adverse Supply Shock"
In supply constrained markets the bidders set the price floor, not the manufacturer. A manufacturer produces a supply; they tender the supply for bids. The highest bidder gets the opportunity to purchase the supply. Everyone else has to wait for the next bid, the waiting is their opportunity cost, and is priced into their bid.
Theres machine learning algorithms in the mix, and it all works perfectly smoothly, until it doesn't.
Smoothly here, being defined as the markets emotional feelings about it all, not, the price went up.
The price went up because it could, as you pointed out.
I agree completely.
It failed, and I feel we are all the better for it. :D
There are some wrinkles to this. If CXMT's RAM was somehow inferior, then the market could sustain two different prices for two different products. But it's difficult to separate the market like this without just making unusable trash. The one separation we do have - DDR5 vs HBM - works against the consumer's favor. One layer down in the supply chain, RAM manufacturers have to buy silicon wafers to etch circuits into, and HBM takes up treble more wafer space than DDR5 per gigabyte. In fact, this is specifically the scarcity that allowed Sam Altman to engineer a RAM crisis, by buying up a bunch of wafer supply that he could then redirect to more AI-optimized HBM.
We can also infer from all of this that CXMT - while producing lots of memory - does not have the scale necessary to actually quench the inflated demand. Either because their yields are shit, or because China wants to run a business and made the same calculation as the other RAM producers that the AI bubble would pop before they could get additional equipment online to service demand.
>Only to see them all end up on the surplus market, unused, for a couple hundred ?
You realize to people that aren't you, this looks like holding two competing ideas at once.
And the comparison makes zero logic. Your library, government, employer are on the internet, who gives a fuck what dump their first modem went into
Lots of business owners here too that build their own clusters
2/ If it's a boom and bust cycle as he said, where's the bust? It's not coming. Humanity will use exponentially more memory and computing, as we've seen. That's not going away.
Let's have interesting conversations about the next steps for memory, is it quantum? what's up next? is humanity really beholden to a monopoly of private chip makers
Why is so hard to get self-hosted bare metal which uses all of the modules? Most server providers go for 1 stick and call it a day....
ffs
This boom is only superficially like the dot com bubble.
There are absolutely parallels.
But, a GPU that can help me write emails, code, articles, etc. has intrinsic value that a Cisco router simply didn't provide.
I don't believe GPUs are a new asset as Jensen Huang claims. (His take seems very bubble-ish.) But, I feel these elevated prices are going to last until production catches up.
If you're writing emails, you're gonna need to send them through a router or two before someone reads them.
> I don't believe GPUs are a new asset as Jensen Huang claims.
He sells GPU's, so taken in context, he's promoting the financing of GPU's, by anyone who is listening to his sales pitch. He travels in CEO circles, so, he's talking to other CEO's, and the banking industry.
The phrasing of it also directly pushes back against the articles accusing the AI industry of circularly financing itself to inflate everyones balance sheet.
And the other thing this idea of it being an asset that you invest in, like a house presumably, is that it primes all of us for 2 future realities: price increases and minimal performance gains ( "its useful for longer so finance it" ).
As to economic value proposition to any individual or group, it boils down to how they are going to use it to recoup the expense.
Don't just take the sales guys word for it, ever.
There have been many videos lately, on whats happening in the automotive market. The industry has been adding more and more tech, to the point where its now financially extractive via mandatory dealer interaction aka fees, and people are becomming fed up with it all, wanting something simpler for less money.
Translate this to the big fat GPU card industry, and how long is it before AMD/Intel add just enough AI ops into every CPU that noone cares anymore and they start asking themselves "Why am I buying 96GB of RAM and a GPU, and 96GB RAM and a CPU ? I'm buying ram twice ! It's a conspiracy !"
Unified architectures seem to be inevitable, GPU's would no longer need seperate ram, then shrinkflation is gonna hit your video card, and you're gonna pay GPU+RAM prices for just a GPU. The upside to this is that you could fit more GPU's in your box and they won't need 2.5 slots each. For now, they are hoping the DGX Spark type boxes fill the market demand, and allow them to sell higher priced products. Unified architectures would dent demand for those.
Just my 2 cents on it all. Enjoy the ride.
That’s what the cable company wants you to believe. We sent their damned boxes back. They wanted us renting a modem that never hangs up, then buying a router just so our own computers could talk to each other. Mine used to call yours directly over the telephone!
Now there’s a computer inside the modem, another pretending to be a router, and a mess of switches, gateways, and subnet masks to do what two modems and a telephone number used to accomplish.
They took the party line, put it on a television channel, called it broadband, and convinced everybody to rent the equipment.
In my day, when somebody tried to break into your computer, you could hear the bastard dialing. And when you were done with the Internet, you hung up. Now they want you connected twenty-four hours a day.
Just wait until somebody breaks into your television through that cable. You’ll wish you’d kept your rabbit ears.
That's not what a router does. A router provides communication between two (or, theoretically, more) different networks.
Multiple computers talking among themselves would generally all be on the same network. (Or, if they're not, you'd put a whole mesh of redundant routers between them.)
I'm sorry, what?! Communications is a valuable thing. How is all that AI slop going to hit the Internet if there's no routers?