There are no dark GPUs. Compute translates directly to money for these frontier labs.
I think everyone is reading way too much into this. Sure there is some circular transactions that are sus, but this ain't it.
I want to make a comparison with a car rental business and say that it would be like valuing Hertz entirely on the basis of the number of cars they own, as opposed to how many they rent out, but cars have a much longer depreciation period, if there are no customers they’re not costing you more money, unlike your computer which you are using for training and sucking up massive amounts of energy, and those cars do maintain decent value even after they’re of little use to the car rental company, unlike the compute here.
The frontier labs are shifting from pricing grounded in the price of compute, to pricing grounded in the intelligence provided, or more specifically the economic value of that intelligence downstream.
The margins on that allow them to pay a hefty premium on compute and still come out ahead.
As they buy more compute at high prices, they're also pricing out competition from cheaper models. It's already become materially more difficult to get compute to run open weight models at competitive prices as a result of frontier labs in the last year.
I think OpenClaw created a mania that was completely unfounded (Apple Silicon is worth dirt compared to literally anything from NVIDIA including consumer GPUs), but the prediction of compute becoming scarce was correct
Opus 4.7 has all the signs of a smaller model distilled from a newer pretraining run... except a smaller price.
Flash 3.5 raised in price pretty meaningfully over Flash 3
GPT 5.4 got a small price bump over gpt-5.3-Codex/gpt-5.2, then gpt-5.5 doubled pricing over gpt-5.4
Even open weights isn't immune: Kimi K2.6 was originally priced higher despite openly being 2.5 + more post-training, same with GLM 5.1 vs 5
-
All while rental prices are spiking month over month, and NVIDIA Inception discounted prices for buying are higher than undiscounted prices for buying 6 months ago...
In the medium term, everyone ramps up production. Huawei and other Chinese companies work really hard to develop in-house alternatives. At some point, the hype cycle will peak and less money will flow into datacentres (yes, this will happen. It always does. Even for technologies that change society. The bubble always bursts).
The question is not if this will happen. It will happen. It's just a question of when it happens and how big the magnitude of the cycle is.
There's a reason old 3090's went from $600 in 2022 o to over $1K in 2026.
How someone can look at an asset class thats appreciated an order of magnitude in the last two years and say it will depreciate in value when the tailwinds are even stronger now is beyond me.
…but the real question whether you want to undervolt your asset if you’re renting it out is why bother? You probably expect to replace it anyway after it’s spec lifetime, for sure want to replace it when a more efficient solution is available since datacenters are power and volume constrained and customers care about performance much more than hardware longevity (otherwise they’d buy instead of rent).
Just waste more money! It's easy.
Since this entire sub-thread is in the context of used 3090s or consumer GPUs in general, you've failed to bring up anything relevant yet again.
Here is your strategy:
1. Increase power consumption by 50%: This costs you more energy to run the GPU, it also costs you more energy to cool the GPU, it ruins the GPU and since you hit power limits of your infrastructure earlier, you will have fewer GPUs in total.
2. Increase maximum performance by 10%: This is hardly noticeable, since the standard inference use case primarily involves taking advantage of the high memory bandwidth of a GPU. This means prompt processing will be 10% faster, or maybe your segmentation model that ingests video runs at 33 fps instead of 30 fps. You're optimizing for winning a benchmark with what will be used hardware in the future, that's asinine.
3. Throw away old GPUs or sell them for peanuts when they still sell for $1000 on the used market if they are in good condition and for $400 if they are damaged. I think the mistake here is obvious. If your GPUs are sold for peanuts, it's because you didn't take care of them.
Your business strategy is obsolete and based around the idea of pre COVID excess hardware capacity before there was massive AI demand where throwing out hardware made sense, because Moores' law was in full swing. Even Google is still offering their v2 TPUs from 2017 even though they've been long since obsoleted. Now in 2026, there isn't enough memory for consumers and people are snatching up all the hardware they can get their hands on. There were some big initial energy efficiency wins from implementing smaller data types that are no longer possible now that fp4 is the smallest possible floating point type that still makes sense and even if you go smaller, you can go down to two bits at best. The parameters are starting to become so small that 2:4 sparsity is becoming unattractive, because it adds one bit to the parameters.
2:4 sparsity for fp4 means 4+4 bits are compressed to 4+1 bits, but 2 bit parameters mean 2+2 bits are compressed down to 2+1 bits.
If you understand even a little bit about hardware, you notice that the tensor core hardware has already been optimized to the extremes and that there isn't much more you can pull out of it. Unlike CPUs there is hardly any control flow in matrix multiplication. The tensor cores implemented in Nvidia GPUs might be a little bit less efficient than an NPU/TPU based implementation (think Google), but there are no more obvious micro architectural improvements here. With CPUs the micro architecture has become so complex, that there may be ways to increase performance further, but for GPUs and NPUs, there is not much left other than process scaling. Further gains require better manufacturing processes from TSMC. TSMC introduced 3nm in 2022 and only started producing 2nm in 2025. That's a three year gap where barely anything happened and all the gains came from going from bf16 or half precision floating point, to fp8 and fp4.
Burning through hardware at high power consumption and mediocre performance increases is clearly not the way to go.
* except ram
The same argument you’ve made would work for tulip bulbs, dotcom prices, or whatever. Prices go up until they don’t. Exponentials don’t last forever and the intrinsics of technology assets depreciate: things wear out and are also replaced with better things.
At some point the market will be saturated with supply and prices will come down for older gen hardware. It can take years though, but it happened to fiber cable and fiber doesn't even depreciate like chips.
the comment you replied to is word-by-word what people hyping canadian telecoms were saying before the dotcom crash!
That's the default assumption but in the new GPU+Memory constrained age isn't true.
Time on 4 year old H100 servers costs more now than when they were new (!!)
Is it an age or a temporary situation?
The key question is on direction of LLMs. Right now, LLMs are taking over human jobs. If the cost of silicon+power < cost of human being doing the same work, what rational reason is there to employ a human being?
If this applies to SWEs, lawyers, business analysts, many research scientists, .... this situation could persist for a long, long time. While capital costs less than the inputs of labor (nominal food, housing, etc.), there is no need for labor.
The key question is about continued progress in models, and of the tooling around them:
- Plateau: Old silicon obsoletes in due course
- Rise quickly: Old silicon maintains value for a long time
To maintain a functioning society and social contract?
Is wanting low unemployment in our society not rational?
However.
It's not rational relative to the short-term incentives of a typical corporation or investment vehicle. PE, VC, fund managers aren't paid to give a fuck about the social contract. Literally not in their job description.
If you want to complain about selfishness then do it on selfish individuals, which by the way, are present in all types of economic systems.
Capitalism provides a set of incentives that shape how people make decisions. Anyone can be selfish, but selfishness in capitalist society has a particular shape. To ignore the external incentives when looking at human behavior is horribly naive and shortsighted, but is frequently done by capitalism-apologists who seek to disregard any criticism of their favorite incentive system.
Only conditionally on there being bad consequences for high unemployment.
I don't particularly trust politicians, but there's a whole host of hypothetical scenarios about futures where work is essentially optional. Unfortunately, they're all either in the sci-fi or religion sections of the book store:
Despite people occasionally investigating UBI, the efforts to research UBI seriously have the same problems that Marx had with literal Communism, in that there's an obvious difference between any partial transition as compared to a global transition, and we don't have a completely disconnected parallel world to be a petri dish for us to test the economic outcomes on.
So okay cool you don't need people to design and build cars. Who's going to buy the cars and where exactly are they finding money?
But see also the "radiologists driving to work" meme for why I think tech in general is currently getting high off their own farts.
Like imagine there was something you could buy where you insert some lumber, give it some passable description of furniture, and it outputs it. And you paid $20/month for access to this. And this was all being bankrolled by the furniture industry? I mean, sure guys - it's much appreciated, but I don't think I've ever seen anybody so enthusiastic about digging their own grave. I think it's already obvious that the gazillion dollars of API calls isn't going to materialize - it seems the handful of companies that trialed that are already reversing course hard. And in the future where LLMs are successful, that'd be even more true.
Both of those are devastating for their valuation. Stopping growth means open modes catch up in a year or so. Continuing means end of the current economy.
The biggest practical issue will be that if these robots ever did start replacing people on a large enough scale, then you'd have a lot of angry, desperate people with a lot of time on their hands. So that alone will probably work as the primary mediating factor.
Quite an interesting time to be alive because the future is so completely impossible to predict. The world just a decade from now could look entirely different, or it could just be self driving cars all over again.
Short term, money physically exists and gets spent, so if you wave a magic want of oversimplification and transition all labour to AI instantly, all the money currently in bank accounts and wallets gets spend on the same businesses it was already getting spent on, a lot of which gets spent on stuff from other businesses who have in this scenario also replaced all their labour with AI.
Eventually, perhaps quickly, all this money ends up in the hands of shareholders and landlords. There's a lot of both in the economy; famously retirement funds, but smaller-scale shareholders and landlords also exist. I wouldn't want to guess what the distribution looks like, probably highly variable between countries not just social classes (the definitions of which themselves can vary between countries).
Long term, money exists as a convenient fiction to help us organise transactions of goods and services: while it may be physically possible to eat gold and banknotes, you're not getting any real nutrients out of it when you do. So in a world where goods and services come from machines, the options are too broad to forecast: humanity could be relegated to the same role and economic stature as other primates (both in and out of zoos), or we could get universal UBI denominated in machine labour credits which lets each of us live better lives than the most extravagant billionaires live today.
I'm just trying to understand if suppose you have fully robotic farms and fully automated slaughterhouses and fully automated McDonald's, who is McDonald's selling anything to and how do these people supposedly buying fully-mechanized burgers have jobs? Something just doesn't add up about this in my head about how this equation balances.
UBI ultimately seems like socialism with extra steps. Mostly is comes across as billionaires desperately begging for an alternative to being nationalized.
Well, people need to eat, so either the customers are on government support, or it comes from passive income, or from savings.
The people without those options, do it the old fashioned way: pick berries, throw rocks at animals, rub sticks for fire to cook them, or starve. Mostly starve, as the maximum number of humans who can survive as hunter-gatherers is 100-1000x smaller than the current global population.
> UBI ultimately seems like socialism with extra steps.
I agree. It's very much "from each according to their ability, oh wait we're all strictly worse than machines I guess that's from each nothing, to each according to their needs".
> Mostly is comes across as billionaires desperately begging for an alternative to being nationalized.
Perhaps, but that feels like claiming they're playing 5D chess, when Zuckerberg only plays Settlers of Catan with sycophants who let him win.
> how do these people supposedly buying fully-mechanized burgers
(Only answer I can think of is political ads).
Industrialization allowed people to shift human labor from agriculture to factories and such.
Seems like intellectual labor became more possible as people looked beyond subsistence but also more valuable since a greater population could drive demand for more than just subsistence related activities.
If both aren’t done by many humans, what’s left? Sports training and massage therapy? Sports training might not even be safe…
OTOH, my current lifestyle is already weird if I think about it. Developing software for a machine that I cannot make myself, whose raw materials I cannot obtain, using energy I cannot produce on my own — somehow entitles me to get a particular amount of goods and services from others including food, healthcare, entertainment, landscaping, and manufactured goods.
We live in interesting times…
Peacock tails.
As in, things where the effort itself is the point, to show off that you are capable of surviving when you consume resources so extravagantly on something other than (or even detrimental to) mere survival.
This includes stuff like hand-made art, being in a literal cult, extreme sports, and also refusing modern medical interventions/seeking out infections.
You may ask how someone can get paid for those things; I don't know, but we did manage to monetise talking to each other (ads on Facebook) and being locked in a house with some strangers while everyone's under surveillance cameras for a few weeks (TV show Big Brother).
I assumed the latter and therefore that the memory is depreciating along with the GPU cores it's soldered onto PCBs with.
... or is it a different argument being made, perhaps that depreciation for GPUs has slowed because rising demand will keep them in service longer?
Google is still running 10 year old Tesla T4s at full capacity.
This is way beyond the expected lifetime.
The GPU shortage looks to be even longer lived.
It's the new normal, get used to it.
The MAG7 isn't pumping all their FCF + new debt issuance into DC's just for fun.
The world is seemingly moving into a era where compute is becoming expensive and scarce.
Only thing that can possibly change this is LLMs hitting a vertical unscalable wall.
More AI compute = more CPU, memory, storage needs.
What I am wondering though is how long can you run such a system at basically full load without interruption before it starts to just physically degrade.
If I have a H100 and I let it run for 4 years at full throttle does it still have the same theoretical value as it had at the start or are the chips just burning out.
I think I remember that back when the cards used for crypto mining were sold en masse on ebay the advice was to stay away from them because they are more likely to fail?
https://www.tomshardware.com/pc-components/gpus/datacenter-g...
Others say that moderate load means a lifespan of ~5 years
Not sure what that means but I would assume that a datacenter will start replacing a node once the error rate hits a certain threshold without really investigating why it failed, so the practical lifespan may be shorter than 5 years even if it would technically still be usable enough
That's because the rate of improvement in silicon manufacturing has been continually declining for a few decades, which has a compounding effect. Just compare the technological improvements in successive decades. 1976->1986->1996->2006->2016->2026.
That's why "in real terms" performance has only been very slowly improving if you compare apples to apples (and not e.g. apples to oranges by reducing precision, like nvidia tends to do, or by comparing chips with x W to an MCM with x*2 W and saying the latter is much faster). The "just halve the number of bits in each generation" strategy has also run out now, there's no more bits to halve.
Temperature is a big factor, as well as current density.
But there's also the # and magnitude of thermal cycles (which translate into mechanical stress, leading to metal-fatigue like effects on contact points etc), attack from chemicals in the air, cosmic radiation, ESD damage & more. Some may matter, some not.
That's why "new" > "used" in case of electronics. Especially since you don't know the (ab)use history of used parts.
Let's not mix up depreciation of real value vs USD price (which is arbitrary, plus government controlled)
There are several confounding factors.
We’ve seen massive inflation since then. So some growth in cost was expected.
More importantly, the current Tech industry almost always starts by selling things at a loss. The increased cost could simply be the industry choosing to not subsidize that particular service anymore.
But also, I don’t think that’s a realistic comparison. Rented out GPUs are likely not a similar use profile as compute used for training LLMs. The latter is likely closer to the cryptocurrency GPUs that are running at full tilt 24/7.
And those things physically burn out.
This is untrue.
H100's are used for training (well were, but are now outdated because B100/B200s are much faster).
Most of the reason people rent H100s is for smaller training runs.
If you are doing inference you usually buy managed capacity at Baseten or something, and that is often priced differently (although it comes down to an extra margin on longer term H100 prices basically).
Inference utilization is often actually higher than training now because so much effort has been spent on optimizing that stack.
There is depreciation, which is taking the purchase price and dividing it across N number of years (typically 5). That's the D in EBITDA and is mostly used as a profitability calculation.
The depreciation of a GPU also gets mucked up in the current GPU financed market as well. DDTL loans. The people running the GPUs often don't even own the GPU, they lease it, so there is nothing for them to depreciate (D).
The analogy that a GPU is like a used car makes zero sense. There is no oil or tires to change on a GPU. They don't wear out in the same way that a rental car would. They are housed in climate controlled locations with clean power. They just don't fail the way that is portrayed in the press.
Useful life of a GPU is based on profitability. When does opex cost more than profitability?
Some companies, like mine, also have support contracts. Anything goes wrong with the GPU (or any part of the system), Dell comes and fixes it at no extra charge. We just migrate customers and workloads to hot spares while the parts are replaced.
As for compute going down in value... the 122TB of enterprise nvme and 2GB of ram in each server that I bought 2 years ago is now worth vastly more than I paid for it. I'm also renting my GPUs out for more money now due to supply being so tight and demand being so high.
So are you using the computers or not? I'd argue that if you're using them for training, then it's not wasted capacity. And if you're not using them, then you can turn them off, so you're not sucking up energy.
I don’t know but this dude at my son’s school has a 32GB RTX 5090 and it’s worth more than what he paid for; and he did the same trick with the RTX 4090 before that.
Until shortages are the rule, these assets are appreciating
Same with GPUs. There is also a huge market for used GPUs from 1-2 generations ago. The A100 is a six year old chip at this point and is still running strong, especially for inference. Like cars, chips can be refurbished and repaired. A hyperscaler or even mid level player here isn't going to hold onto chips for their entire usable lifespan.
Let us pin this comment and see how it ages
Nothing about this deal is about better technology or talent. It's about an opportunity that's too juicy for Google to pass up on.
This might not be true. Someone was comparing Nvidia's production rate with known data center capacity, and they do not match. Their conclusion was that people (possibly even Nvidia) were hoarding GPUs- in the very short term this might be a good strategy, but GPUs go EOL fast. There are other stories about paused datacenter builds that match with this.
TSMC is definitely fully allocated, based on current 40 wk lead times for FPGAs..
This is a reference to the 1990's dot com bubble where internet infrastructure companies overbuilt network capacity, leading to the term "dark fiber". That was an indicator of a bubble because it showed that capacity was larger than demand. OP is saying that this is specifically NOT happening in the case of GPUs yet, indicating that demand still outstrips supply of compute.
>GPUs go EOL fast
We are seeing the opposite of what was expected, GPUs are actually getting more valuable because demand is so great, something that basically never happens. Even older chips have become more valuable.
>paused datacenter builds
It doesn't seem that datacenters have been paused because of lack of demand for AI, it seems mostly that there is a lot of pushback by cities to build these things and also there is a shortage of power to run them.
IMO none of these things point to a AI being a bubble (over-hyped, demand does not match the stated value). It mostly points to the opposite, there is massive demand for AI and every layer of the supply chain is struggling to keep up with that demand.
The physical world can’t be patched overnight, and cutting edge manufacturing takes a long time. Fortunately we are in a very peaceful low tension world right now and no one would try burning down or blowing up one of those extremely important, irreplaceable fabs.
Huh, anybody want to buy a GTX 680? Or even a formerly-SLI'd pair?
I agree the demand is there, but hyperscaler capex is what now? 3% GDP? This is an absurd amount of money and people who question whether the ROI is there have a point just because of the order of magnitude of this spend number.
Yes, the demand is there for the currently unsustainable price. Lets see what happens when the dumping of money into AI stops and the companies are forced to increase prices a lot.
To reap massive profits before depreciation is just plain smart. LLM space, model generation is just plain crowded now too. And everyone thinks a crash is coming.
They could also build out their own end-user infra, but letting someone else which already sells direct to the public do so, is sensible.
I know of the desire to show profit for the IPO, but my point is, this is a good move on its own.
The NVIDIA GPUs, HBM, land-use permits and power-supply agreements xAI nailed down are absolutely not commodities.
I think xAI is a mess. But let’s call a spade a spade, they speculated on AI compute and they are currently right.
Don't you mean gas turbine purchases and questionably legal operation? But yeah I feel exactly the same way. The AI part of xAI looks like a mess but it seems that they still managed to score a massive win.
The point is it’s running. They built fast before the backlash got organized. Now everyone has to deal with delays and thoughtful permitting processes.
Its not that there isn't value in that business, but it's not the AI business either. Its the one where Oracle is laying off staff to try and avoid a revenue crash on future commitments.
Both Google and Anthropic would be trying to can this sort of rental arrangement as fast as possible since it's a mind bogglingly expensive way to get something you already do in house.
The "backlash" is the poorest residents one of the poorest large cities in America trying to fight for their right to clean air.
Your point might end at "it's running", but holistic thinkers have no problem considering the how they arrived there, given what it's doing to these folks for marginal benefit.
It's not like xAI would go under if they had chosen a less populated location and waited to get permanent power.
Sorry, I'm referring to the national pushback against datacenters being built in peoples' backyards. xAI didn't face backlash. At least not organised enough to stop them. Their competitors, today, are facing backlash sufficiently powerful to stop new datacenters from being put down.
If they were speculating on compute, it seems highly unlikely they'd have spent the operating costs for the last 3 years of model development and deployment instead of just getting even more compute.
It's sheer brute force, tons of waste, seems like very little thought going in to fitting the implementation to the problem.
The value of compute can drop significantly in the event of users figuring out how to optimise for their particular need. And yep, there are wasteful applications that can burn whatever compute is available, but how much demand for that is there when it's properly priced?
Extreme example. Generating novel 4K VR video on demand. I'm certain there's a market for it, at $10/hour probably quite a healthy one, at $100/hour not so much.
Sundar Pichai at Q4 2025 earnings call: “We’ve been supply-constrained".
Satya Nadella, 2026: Microsoft would increase total AI capacity by over 80% in the year and roughly double total datacenter footprint over two years.
Microsoft CFO, 2026 earnings call: “We’ve been short now for many quarters. I thought we were going to catch up. We are not. Demand is increasing.”
So yeah, either top management of hyperscalers are doing a 'bit' for the last few years, or Aschenbrenner 'Situational Awareness' is going roughly as predicted and hyperscalers are desperate to acquire compute even at higher cost.
There are actually lots of GPUs in storage somewhere waiting for data center megawatts to put them in.
In fact, for all these companies to do what they're going to do, they need a massive, massive massive amount of data centers, a highly improbable number of data centers that need to be built in an highly improbably short amount of time.
And the capitals about to dry off in about a year. So it's a race between these improbable timelines on data center construction, with capital evaporating.
- Ukraine war similarly is triggering an EU buildup and reduction in us dependency
- all the IPOs indicate the companies themselves know the private investment is coming to an end so they need the retail investors to keep the boondoggle moving
Alphabet/Google profits:
Q1 2025: $34.54 billion
Q2 2025: $28.20 billion
Q3 2025: $34.98 billion
Q4 2025: $34.46 billion
<<Q1 2026: $62.58 billion>>
Amazon profits:
Q1 2025: $17.1 billion
Q2 2025: $18.16 billion
Q3 2025: $21.2 billion
Q4 2025: $21.19 billion
<<Q1 2026: $30.3 billion>>
Both Alphabet/Google and Amazon have invested recently into Anthropic and are doing all sorts of financial chicanery.
https://www.youtube.com/watch?v=-bjNrGFiAI4
Nah, man, it's all fine, they're just going to take down the entire global financial system doing this crap, and by global, I mean <<everyone's>> pensions are going to take a hit, even "fully funded" pension systems.
bko didn’t say there isn’t circular financing going on. They’re just saying this isn’t an example of it. They’re right.
It’s a potential conflict of interest. And if the agreement is fake—if Google cancels without paying the cash—it could be market manipulation. But the influencer space likes to latch onto jargon, and the one it’s overapplying right now is circular financing.
What are you even going on about?
[1] They're indirectly tied to it.
They were unrealized gains on non-marketable equities. It’s clearly disclosed and done according to GAAP. It’s put under other income precisely so analysts can strip it out when modelling long-term trends.
Like, yes, if SpaceX goes to zero Google would have to realize losses and probably lose a quarter or two of GAAP profits. (But not cash flows. Cash-flow wise, it may wind up being positive due to tax effects.) It’s a risk factor, of course, but far from making no sense.
None of which is particularly relevant to the deal at hand other than in raising a potential conflict of interest among related parties.
When I said "it makes no sense", I didn't mean "the accounting math doesn't work out". I meant "raising a potential conflict of interest among related parties".
This whole AI financing this is the motherlode of "potential conflict of interest among related parties".
And people who are obtuse enough to ignore this because it's not illegal right now will discover 5-10 years from now that laws are written in blood (or massive bankruptcies).
Sure? Lots of things are potential conflicts. In the Google and Anthropic deals, I'm not seeing evidence of problems.
And that's saying something, because we have a lot of evidence of actual circularity or related-party deals being done with no arms-length anything across AI, in many cases in ways that definitely do see like they are illegal.
And that's where my video comes in. Google and Amazon are very likely juicing up their share price. Of course, in this day and age we can't prove it's a pump and dump anymore...
Google itself has a good reputation as a facilities operator. SpaceXAI is operating gas turbines emitting exhaust at ground level.
- https://cloud.sustainability.watch/explore-issues/example-go...
- https://www.sfgate.com/national-parks/article/mount-hood-wat...
They also seemingly dropped their net-zero climate goal:
https://www.tomshardware.com/tech-industry/google-quietly-re...
All this investment is completely driven by the companies leading the pack. OpenAI and Anthropic have been telling everyone they need to spend hundreds of billions in a few years. Of course they don't, they could do this over 10-15 years and still be profitable. But they're terrified they won't be able to dominate the market. So to dominate the market, they've estimated they need this growth to beat China (and each other). And the US technically has the capital to make this happen, but there's only so much money available to spend. By growing too fast, they spend money faster than they can make it, and the bills are so big that the investors go bankrupt.
That's what happened in the panic of 1873 (railroads instead of AI). That's what's going to happen here in the next 2-4 years.
but it's really bad news for the industry capacity if your best option is unproven space datacenters.
"you have compute, i need compute, i'll pay you for some compute.".