"Secondary market data shows moderately-used 2 to 3 year old GPUs trading at 50% to 70% of new pricing under normal conditions."
That's for good NVidia H100 units.[1] There's a shortage of those. That seems to be the price after removal, cleaning, testing and refurbishing. Raw units removed from a shutdown will not be as valuable.
H100 units are available on eBay, but multiple sellers are using the same picture of a new unit in its original packaging, a bad sign.[2] Some even have pictures with the logos of a competitor.
Who is buying these at 70% of new pricing given the sky high likelihood of them being shot? Maybe it's safe to buy from small labs that went under quickly, but I can't imagine a cluster that has been operating near its thermal limits for a couple years fetching that kind of resale.
It was true of crypto GPUs too, although mostly people picking them up for gaming. Always seems high to me too but if you can get any guarantee of them not being on fire when they were pulled the bathtub curve keeps you pretty safe, thermal limits are limits for a reason.
Most crypto mining on GPUs would use 100% of memory bandwidth, but only a fraction of the compute available. This is a consequence of ASIC resistance of their mining algorithms -- custom silicon can only offer a modest benefit over GPUs if the hard part is memory bandwidth.
I half expect Nvidia to have buyback contacts like Ferrari with the larger customers to prevent a price crash when they all upgrade and to keep them scarce.
I hope not though, perhaps I can pick up a H100 in a few years if they get sold on the open market.
I worked at an org that had a substantial on-prem GPU datacenter. We transitioned to <Big Cloud Provider> with a substantial negotiated discount rate, with part of the contract being we would sell them all of our hardware and not purchase any more.
If you're planning to run in clouds, committing to not buy hardware (during the contract term, presumably) isn't a big imposition. Maybe you switch to a different cloud, and you wouldn't buy hardware for that.
If you want to switch back to on prem, there's probably a way to structure acquiring hardware so it doesn't break the contract. Maybe you lease it, maybe the purchase happens through a related company, maybe there was no way for the contracted cloud to find out...
I don't know if this was IT shrugging me off or if it was something real but some IT person at this big ISP I worked at told me that they cannot just buy an SSD — my windows box at work was running off of a hard disk in 2019 — and that there was some contract that said any computer hardware we bought had to be through HP or something like that and it takes many months it something like that.
The usual very short term corporate thinking that maximizes quarter profits while bankrupting the company in the long term.
IMO a very shortsighted decision, if not downright stupid.
There's going to be a golden age of GPGPU compute in the next few years once A100/H100 are fully obsolete for running frontier models efficiently and the price plummets
It will be perfect for stuff like GPU-accelerated query engines, "classical ML" and every other CPU-based workload that could conceivably be offloaded to GPU
GPGPU? General Purpose GPU? If embarrassingly parallel CPU algorithms weren't offloaded to the GPU previously, why would the A100/H100 price drop make a difference? We had cheap GPU in the past and we still left plenty of performance on the table with CPU programs because they were easier to build.
Is the idea that previously maintaining GPU programs was expensive whereas now AI makes it cheap? If so, I could buy that line of reasoning.
Maybe relatedly, I expect (hope) the hardware manufacturers will ramp up supply in the meanwhile which would also put downward pressure on GPUs. Right now though this hardware crunch is making me sad, not even because of GPUs but also because of general memory / disk.
The only reason a datacenter would ditch their H100 is if it becomes uneconomical to run them, with newer silicon providing much more power efficiency. When that happens they'll look like a used V100 looks today: horribly inefficient, lacking modern data types and engines, requiring screaming server fans with weird adapters to not melt, way beyond end of life in terms of cuda support. Almost completely damn useless unless you really have no other alternative.
It’s common for car companies when they enter a new market. It removes uncertainty from the second hand market.
By doing that, you know upfront what the value of your used hardware will be at the time you decommission it. It removes a lot of the risk for buyers in a volatile market.
so are buybacks. you choose to sign the contract. there's no way they didn't have an escape clause, although likely it meant not using the cloud provider anymore
Yes. "You choose to sign the contract" is not a reasonable argument, for two reasons:
1. There is one supplier, so you have no choice.
2. Even if you had a choice to sign the contract, this still means that it's not the same as a trade-in, because trade-ins are always voluntary, but once you have signed the contract, a right of first refusal is not.
In general, the "you chose to sign the contract" argument is a poor justification for bad contracts. If the contract is bad, it is bad regardless of whether you chose to sign it.
They're pretty specific to the datacenter use case with no outputs and they need to be cooled externally, principally through the very loud high speed fans used in data centers. I suppose you could strap a fan to one and put it in a normal case or maybe make a dedicated after market cooler (like the water blocks made for water cooling cases).
I think there's a market for a home AI server that can run an LLM or video gen model behind a web frontend. Not literally a raspberry pi strapped to an H100, but something with lopsided enough specs that people joke it is.
And precisely because it's such a huge headache to do yourself, I think a small company could make a nice business wrapping up used datacenter cards in that sort of server.
My workstations are usually tower servers, which are the same design as their 4 and 5u rack counterparts.
Until the thermal management kicks in, they sound like jet planes. When the thermal management starts and assesses the required cooling, it’ll throttle down the fans to reasonable levels.
That is, until the moment you push the machine to its limits. When then happens, you might get back to the same levels of the boot time, but it’ll require you to push everything to the max - CPU, memory, storage (all 24 bays) and so on. For a normal user, there is a lot of room and it’s virtually impossible, even with a dozen of Teams windows open.
The GPU alone has a TDP of 700W, together with everything else (CPU, RAM, storage, fans) you're looking at 1500W+. Depending on the country, that may be enough to saturate your home's electricity uplink...
That's actually less bad than I assumed. High-end gaming GPUs are already almost at 600W with peak usage above that. I assumed it was much worse than that; that seems absolutely feasible for running at home.
Never heard electrical uplink before but to those wondering Italy, India, and Japan all have requirements around 3kw. I was slightly surprised by this, but in the end if your running this in a tiny space with that low of power your already probably not buying used H100s.
In North America we are on 120V, making a standard 15A outlet only 1500W max, and something like 1200W sustained. To use higher wattage appliances, we have to upgrade our outlets to 20A (2000/1600W) or up our voltage to 240V, but that carries a different set of plugs and outlets as well.
Usually the home service is 200+ amps these days which is 24 KW total across all circuits, though you're right that most individual circuits are only 10-15 amp.
In the US they have a 120V system so it might actually saturate a normal socket over there. My PC room has a 16A fuse and 230V though, so should be plenty :)
> The job of an operations team is to keep all of this in steady state. They know which racks run hot in summer, which cooling loops have been flaky since the last firmware update, which jobs to re-route when a node degrades but has not failed yet. None of that knowledge is written down. It lives in the team.
Hmmn. All of this information should live in the monitoring system, in which case any frontier model will be able to get to grips with it in short order. It feels like the author doesn't really fully understand the changes brought about by the systems they are writing about.
The relatively slow depreciation of GPU value is an artifact of supply constraints. If you run fp4 inference and could choose freely between Hopper and a Rubin, the performance per watt would make the Hopper unattractive even if you paid zero for the hardware and only for the power.
You can't get the Rubin, or even the Blackwell, so you will pay for the H100 but this won't last if fabs ramp up capacity.
Not to mention, physical limits to lithography are slowing down significantly... so tech will continue to evolve more slowly... it'll never be the jump from 1080-1990 again, for example, even though 1990-2000 was pretty close, 2000-2010 much slower and since 2010 slower still.
What's as or more weird is how much hardware is backordered, and how much live hardware is allocated, but waiting on facilities for operation. And how many facilities are years behind at this point already... all on various credit and dept swaps between all the involved companies... it's not just a balloon, it's a house of cards balanced on a balloon.
The increase in PFLOPS/dollar has continued accelerating, a lot from process, but also a lot by simplifying the architecture- if you had placed an H100 worth of transistors on a CPU-like architecture, you wouldn’t reach the same peak performances.
Nah, "selling shovels" is mostly a metaphor, the amount of iron that went into mining equipment was insignificant, beyond some local demand peaks. The majority of the business was consumables.
A key thing to understand about the gold rush is that it was not a major economic event, or at least nowhere near as big as the participants thought it would be, hence the tradegy.
The AI gold rush is different in that there actually is a mountain of "shovels" large enough to flood the global market quite severely.
>... when I left Paperspace in mid-2024, our M4000 GPUs, nine-year-old GPUs, were still consistently utilized at near-total capacity. That’s not a typo. Nine. Years. Old. Still booked, still working, still generating revenue.
As such you'd assume these cloud providers to want faster depreciation of their GPU assets rather than slower? I suppose in this case they do have an incentive to show bigger revenue numbers, but there seems to be a downside to this that is not being discussed?
https://news.ycombinator.com/item?id=49005798 "I have an M3 ultra mac studio with 512 GB of memory. I want to sell it, [...] Given the high cost of the Mac studio ($20,000 or more)"
Very difficult to sense check that substack post without access to the TLB credit agreement.
There are likely management service agreements from xAI proper -> SPV to cover precisely what the author talks about. Clearly, xAI could play games but without seeing the docs (which are not public), it's very difficult.
This article's basic point is right though. On the other hand, the LTV of this deal was approx 50% debt-financed (not too high; very much depends on the "V"). At 12.5%, it's not as if its being priced as a high quality asset.
Overall, substack post was too bearish. The wider point is that there's a lot of froth tied to what has now become systemically opaque - namely the circular deal flow that every hyperscaler, nvidia, neoclouds and friends are now engaged in. When the proverbial hits the fan, that stuff will be difficult to price and find few willing buyers with the competence to underwrite.
The systemic issues are the bigger concern than one specific deal imo.
All indications are there will be a lot of repossessed GPUs appearing on the market before too long. Likely to be messy for a while but will open up a lot of possibilities when it’s easy to get your hands on some secondhand GPUs.
Isn't it worth exactly what you can sell it for. a few ways to do this.
slow and awkward, best market match: The auction. sell to highest bidder.
faster and more customer friendly but poor market match until a lot of units sold: The store. guess price, adjust up or down to reach sell frequency desired.
fast and good market match but takes a knowledgeable customer base: The reverse auction. Start with price too high lower it over time until it sells.
I like Meg a lot a human, but Meg is all doom and gloom. Every single post she makes is about how GPUs fail [0] and now she's onto how financing is a big thing just waiting to crash and "nobody knows what a used GPU cluster is worth"...
Actually, we do, people offer them to me all the time. A used box of MI300x is $257k. "There is no GPU futures market"... actually there are a few of them that people have pitched to me.
This article is a lot of words from someone who isn't actually buying or deploying compute. My point is... take it all with a grain of salt.
Somehow at any given moment every possible topic to discuss on HN belongs to either the set of "it's amazing and nobody can say anything bad about it" or the set of "this thing sucks and nobody is allowed to say anything positive about it" and I never have ANY idea which one any given topic will be in on any given day.
In 2008, was the opinion of a banker more "insightful" than they opinion of journalists, bloggers and normal people talking about the imminent subprime crash?
latchkey is being restrained. He runs a data center filled with AMD GPUs. He's got a lot more insight to the real, lived experience of such a business than the post seems to have.
This is a good article. I've been curious about how this is going to play out. A couple of data points:
1. An enthusiast had a project to get a V100 working on his PC [1]. This was a ~$10k GPU 10 years ago. It's now sold for scrap;
2. The A100 came out in 2020 and cannot run a large model like DeepSeek v4 Pro. It can run Flash. You need a 16xH100 cluster to run Pro and that's a ~4 year old GPU and AFAICT 8xB100 or 4xB200;
3. We're about to roll out R100/R200s.
I'm surprised that NVidia is moving to a 1 year product cycle (per this article) because the big question I've had is what's that going to do to existing investments in GPUs. Why? Because if 4xR100 can do the work of 32xH100 then that's a massive advantage in performance-per-Watt, which I think is going to be the only metric that ends up mattering.
In addition to raw power, new capabilities are developed and come online. For example, certain smaller, more efficient quantization methods just didn't exist on older hardware.
Oh, another thought from this: a 9% annual failure rate just goes to show you how ridiculous the idea of orbital data centers really is. Orbital DCs were always just a pump-and-dump scheme for SpaceX's IPO.
Currently it gets expensive to run models larger than ~31B locally. You start to need some pretty expensive hardware. That's going to change. I don't expect we'll be running 1T+ models on a Macbook Pro within 5 years (at reasonable inference rates) but I think people today will be shocked at what's being run locally in 5 years and that'll easily be 100-200B+ models.
Is the 30-50% of face value realistic for "Liquidation Value"? If one organization has to liquidate, sure. But if the bubble pops and many groups have to liquidate at once? Owners will be lucky if they can dodge the recycling fees.
Essentially nothing, fractions of a penny on a dollar, because players who could afford paying real money won't risk the crusty old hardware - so you're limited to buyers who still need a massive cluster but don't have AI infinite money glitch enabled
> These are not catastrophic events. They are the steady state.
> There is no GPU futures market, no standardized residual value curve, and no way to lock in a forward rental rate. The premium is is the price of underwriting in the dark.
The headings are also AI like, a lot of essays before usually did not have titled sections but now they do and they all feel like these.
In addition the diagrams themselves look pretty AI generated.
Would it make economic sense to strip it down and sell for parts? That’s how it’s done now for older data centers, where the obsolete equipment is sent off to China, stripped down for parts, and sold on the secondary market. I’ve picked up several older but still useful RAID hardware cards off eBay this way.
I’m mainly interested in getting some DDR4/5 and RTX5090s on the cheap :).
so you're telling me you can't use 1 or 2% percent of 5 billions dollars to rebuild a team that runs GPU clusters for like a couple of years ?! and the guy that borrowed billions from these banks would want to mess up that relationship for what? I mean he is a stupid narcissist but not to that degree. This whole article makes no sense to me.
Given the price for a rack of bc-250 after the crypto hype cycle, the expected value of the hardware will be around 5% to 10% of the original retail price.
Without other market influences, that is a >90% expected discount when the over-provisioned market must inevitably self-correct.
If the Market follows what Samsung/SK Hynix did to the South Korean exchange this week, than the "AI" bubble will hit harder than the dot com crash.
I like the Shrek Movie correlation theory, as they always happen just before Debt-backed investors get hit hard... And the new film is due out in 2027. =3
If anything, you might have to pay to have them disposed of, they don't really have any meaningful used eBay market outside of the randos that want to do high end extreme local inference in their basement.
Also, as for RAMmageddon, the inference SBCs that all of the AI bros bought don't have DIMMs, they're not even the right chip: its all GDDR and LPDDR. The only DDR DIMMs being consumed are for regular non-inference machines that help run the business and service infrastructure behind the scenes.
> Silent data corruption (SDC) is the most expensive, where a faulty GPU produces wrong answers without crashing anything, which means a multi-day training run can complete normally and the resulting model weights are quietly poisoned.
That still doesn't tell you what GPUs will be worth in 5 years, because it depends upon what inference demand is like and what the alternatives are. What will an hour of NVL72 be worth in 2029?
(And other things, that we know partially but not fully-- like what failure rate for current generation parts will be under this loading).
So we have big uncertainties about the revenue, moderate uncertainty about the proportion of the asset that will survive, and some uncertainty about what operating costs will be. It's difficult to turn this into a residual value.
Finally, the whole "operating the big facility" thing is not likely to be plug-and-play for a new technical team following a default. How much outage/disruption ensues?
That's for good NVidia H100 units.[1] There's a shortage of those. That seems to be the price after removal, cleaning, testing and refurbishing. Raw units removed from a shutdown will not be as valuable.
H100 units are available on eBay, but multiple sellers are using the same picture of a new unit in its original packaging, a bad sign.[2] Some even have pictures with the logos of a competitor.
[1] https://introl.com/blog/secondary-gpu-markets-buying-selling...
[2] https://www.ebay.com/shop/nvidia-h100-gpu?_nkw=nvidia+h100+g...
It was clean, cheap, and is still going strong for daily gaming.
In its working life it was undervolted and probably cooled better than in my rig.
I hope not though, perhaps I can pick up a H100 in a few years if they get sold on the open market.
why would anyone sign such a contract?
If you want to switch back to on prem, there's probably a way to structure acquiring hardware so it doesn't break the contract. Maybe you lease it, maybe the purchase happens through a related company, maybe there was no way for the contracted cloud to find out...
It will be perfect for stuff like GPU-accelerated query engines, "classical ML" and every other CPU-based workload that could conceivably be offloaded to GPU
Is the idea that previously maintaining GPU programs was expensive whereas now AI makes it cheap? If so, I could buy that line of reasoning.
Maybe relatedly, I expect (hope) the hardware manufacturers will ramp up supply in the meanwhile which would also put downward pressure on GPUs. Right now though this hardware crunch is making me sad, not even because of GPUs but also because of general memory / disk.
By doing that, you know upfront what the value of your used hardware will be at the time you decommission it. It removes a lot of the risk for buyers in a volatile market.
Right, so they're not voluntary.
1. There is one supplier, so you have no choice. 2. Even if you had a choice to sign the contract, this still means that it's not the same as a trade-in, because trade-ins are always voluntary, but once you have signed the contract, a right of first refusal is not.
In general, the "you chose to sign the contract" argument is a poor justification for bad contracts. If the contract is bad, it is bad regardless of whether you chose to sign it.
Maybe someone could start a business buying up and rehousing these.
And precisely because it's such a huge headache to do yourself, I think a small company could make a nice business wrapping up used datacenter cards in that sort of server.
if you haven't heard a 5u server intended for a datacenter rack come to life it's quite the experience. Sounds like a plane taking off.
Until the thermal management kicks in, they sound like jet planes. When the thermal management starts and assesses the required cooling, it’ll throttle down the fans to reasonable levels.
That is, until the moment you push the machine to its limits. When then happens, you might get back to the same levels of the boot time, but it’ll require you to push everything to the max - CPU, memory, storage (all 24 bays) and so on. For a normal user, there is a lot of room and it’s virtually impossible, even with a dozen of Teams windows open.
These things get hot and are fussy about their requirements.
In North America we are on 120V, making a standard 15A outlet only 1500W max, and something like 1200W sustained. To use higher wattage appliances, we have to upgrade our outlets to 20A (2000/1600W) or up our voltage to 240V, but that carries a different set of plugs and outlets as well.
It's 1800W for short periods and 1500W sustained.
One that runs continuously
Hmmn. All of this information should live in the monitoring system, in which case any frontier model will be able to get to grips with it in short order. It feels like the author doesn't really fully understand the changes brought about by the systems they are writing about.
You can't get the Rubin, or even the Blackwell, so you will pay for the H100 but this won't last if fabs ramp up capacity.
What's as or more weird is how much hardware is backordered, and how much live hardware is allocated, but waiting on facilities for operation. And how many facilities are years behind at this point already... all on various credit and dept swaps between all the involved companies... it's not just a balloon, it's a house of cards balanced on a balloon.
The increase in PFLOPS/dollar has continued accelerating, a lot from process, but also a lot by simplifying the architecture- if you had placed an H100 worth of transistors on a CPU-like architecture, you wouldn’t reach the same peak performances.
but I generally agree, people put a lot of faith in the exponential leaps vs the exponential space.
You tell them we're not living on mars any time soon and they'll bring up christopher columbus.
A key thing to understand about the gold rush is that it was not a major economic event, or at least nowhere near as big as the participants thought it would be, hence the tradegy.
The AI gold rush is different in that there actually is a mountain of "shovels" large enough to flood the global market quite severely.
Which has this anecdotal data point:
>... when I left Paperspace in mid-2024, our M4000 GPUs, nine-year-old GPUs, were still consistently utilized at near-total capacity. That’s not a typo. Nine. Years. Old. Still booked, still working, still generating revenue.
Also, I won't claim to understand accounting, but in general it seems it is advantageous to accelerate depreciation schedules for high CapEx industries because they lower taxes: https://leyton.com/us/insights/articles/what-is-accelerated-...
As such you'd assume these cloud providers to want faster depreciation of their GPU assets rather than slower? I suppose in this case they do have an incentive to show bigger revenue numbers, but there seems to be a downside to this that is not being discussed?
You go to ebay search for a used GPU. You get a price.
Neither is used servers a new thing or used routers. There are established used server companies.
https://en.wikipedia.org/wiki/The_Emperor%27s_New_Clothes
There are likely management service agreements from xAI proper -> SPV to cover precisely what the author talks about. Clearly, xAI could play games but without seeing the docs (which are not public), it's very difficult.
This article's basic point is right though. On the other hand, the LTV of this deal was approx 50% debt-financed (not too high; very much depends on the "V"). At 12.5%, it's not as if its being priced as a high quality asset.
Overall, substack post was too bearish. The wider point is that there's a lot of froth tied to what has now become systemically opaque - namely the circular deal flow that every hyperscaler, nvidia, neoclouds and friends are now engaged in. When the proverbial hits the fan, that stuff will be difficult to price and find few willing buyers with the competence to underwrite.
The systemic issues are the bigger concern than one specific deal imo.
slow and awkward, best market match: The auction. sell to highest bidder.
faster and more customer friendly but poor market match until a lot of units sold: The store. guess price, adjust up or down to reach sell frequency desired.
fast and good market match but takes a knowledgeable customer base: The reverse auction. Start with price too high lower it over time until it sells.
Actually, we do, people offer them to me all the time. A used box of MI300x is $257k. "There is no GPU futures market"... actually there are a few of them that people have pitched to me.
This article is a lot of words from someone who isn't actually buying or deploying compute. My point is... take it all with a grain of salt.
[0] https://x.com/meggmcnulty/status/2040851080066859386
I was complaining that it's obviously incorrect that nobody knows what used GPUs are worth, not about latchkey.
I have NO idea why anyone is upvoting a post titled "nobody knows what a used GPU cluster is worth", that is a WILD claim.
1. An enthusiast had a project to get a V100 working on his PC [1]. This was a ~$10k GPU 10 years ago. It's now sold for scrap;
2. The A100 came out in 2020 and cannot run a large model like DeepSeek v4 Pro. It can run Flash. You need a 16xH100 cluster to run Pro and that's a ~4 year old GPU and AFAICT 8xB100 or 4xB200;
3. We're about to roll out R100/R200s.
I'm surprised that NVidia is moving to a 1 year product cycle (per this article) because the big question I've had is what's that going to do to existing investments in GPUs. Why? Because if 4xR100 can do the work of 32xH100 then that's a massive advantage in performance-per-Watt, which I think is going to be the only metric that ends up mattering.
In addition to raw power, new capabilities are developed and come online. For example, certain smaller, more efficient quantization methods just didn't exist on older hardware.
Oh, another thought from this: a 9% annual failure rate just goes to show you how ridiculous the idea of orbital data centers really is. Orbital DCs were always just a pump-and-dump scheme for SpaceX's IPO.
Currently it gets expensive to run models larger than ~31B locally. You start to need some pretty expensive hardware. That's going to change. I don't expect we'll be running 1T+ models on a Macbook Pro within 5 years (at reasonable inference rates) but I think people today will be shocked at what's being run locally in 5 years and that'll easily be 100-200B+ models.
[1]: https://www.hackster.io/news/hacking-a-server-grade-nvidia-g...
In my experience, AI is easier to read than this was.
> These are not catastrophic events. They are the steady state.
> There is no GPU futures market, no standardized residual value curve, and no way to lock in a forward rental rate. The premium is is the price of underwriting in the dark.
The headings are also AI like, a lot of essays before usually did not have titled sections but now they do and they all feel like these.
In addition the diagrams themselves look pretty AI generated.
I’m mainly interested in getting some DDR4/5 and RTX5090s on the cheap :).
Without other market influences, that is a >90% expected discount when the over-provisioned market must inevitably self-correct.
If the Market follows what Samsung/SK Hynix did to the South Korean exchange this week, than the "AI" bubble will hit harder than the dot com crash.
I like the Shrek Movie correlation theory, as they always happen just before Debt-backed investors get hit hard... And the new film is due out in 2027. =3
Can you tell us more about this? Or some link
If anything, you might have to pay to have them disposed of, they don't really have any meaningful used eBay market outside of the randos that want to do high end extreme local inference in their basement.
Also, as for RAMmageddon, the inference SBCs that all of the AI bros bought don't have DIMMs, they're not even the right chip: its all GDDR and LPDDR. The only DDR DIMMs being consumed are for regular non-inference machines that help run the business and service infrastructure behind the scenes.
Anyone know how these get caught ultimately?
(And other things, that we know partially but not fully-- like what failure rate for current generation parts will be under this loading).
So we have big uncertainties about the revenue, moderate uncertainty about the proportion of the asset that will survive, and some uncertainty about what operating costs will be. It's difficult to turn this into a residual value.
Finally, the whole "operating the big facility" thing is not likely to be plug-and-play for a new technical team following a default. How much outage/disruption ensues?