In order to not un-build the data centers, they at least have to make more than it costs to operate them, and also not have some attractive liquidation value (the land, maybe).

I could imagine something like “inference is done at home or in China, that’s the price to beat” and it’s not worth keeping all those GPUs cool out in Nevada.

The $1T number seems more promises than reality, which is closer to the $300B to $500B level. Still a big number, but between a third and a half of the value used in the popular media.

Chips age and fail with age. You can check hot-carrier injection, bias-temperature instability and electromigration as they are the main aging mechanisms. All if these are a linear function of time but exponentieal of temperature. 90-100C these chips are running at are really tough, so they are likely to fail at couple of percent to 10% range in 2-3 years depending on the margins they have in the design.

The solder joints are notorious to fail at a high rate too.

There are data centers that use and rent out 10 year old server GPUs.

They can't run larger modern models. They can't run smaller models as fast as newer servers. So their remaining market is applications where customers are okay with older, smaller models and slower performance.

They have to price the service lower than competitors due to the lower performance. The older GPUs are less efficient so it costs them more to keep them running. They're paid off, but they're taking up valuable power, space, and cooling in a data center.

Eventually there is a tipping point where it's better to replace that space and power budget with something new that has more demand.

The parts are sold off on the open market. There's an equilibrium demand for the parts from other data centers keeping older servers running and from hobby people who are okay with a jet engine sounding toaster of a GPU running in their home.

In addition to the physical depreciations other comments mentioned I'd also mention that old chips will settle into a low price and then actually go up on a per unit basis if you're trying to buy a significant amount of them. With a limitation on fabrication facilities continuing to pump out older cards is an opportunity cost to the manufacturers that would prefer to be producing newer cards. If you were in a place where you suddenly wanted to buy 10,000 3080s, as an example, I'm not certain if the market could actually fulfill that demand and no one with the ability to increase the available supply to meet that demand actually wants to do so.

Chips do wear out and need to be replaced (entropy do be like that and durability is not a primary concern for chip design) so you'll need to refresh your stock and, even if you don't need cutting edge models, the price of all chips at scale will go up over time. It may feel unintuitive since, when the PS3 was released PS1s were extremely cheap - but if you're struggling to understand this effect from your experiences in the consumer market you're actually looking at the price factor that starts making antiques increase in value since at a certain point they become scarce goods. The market price for an NES is higher today than it was in 2003 because the price had already bottomed out from demand from the general consumer market but the demand remaining (speedrunners and the like) is now fixed or growing while the supply is inevitably shrinking.

I used to work in datacenters, during spinning disk era we had technicians from vendors basically every couple of days to replace some broken part. When the massive switch to ssd happened instead of having them every couple of days it was 3 or 4 times per month.

Despite no moving parts things broke anyway and, even if it doesn't break, the vendor can make you change the technology just by playing with maintenance cost of the older one, limiting or removing spare parts from the market.

Gradually, and especially when hot. Modern chips are pretty close to the physical limits of how small they can be made, and that means atomic/chemical effects like electromigration are accounted for and determine the lifetime. Every extra 10 degrees Celsius of temperature doubles the speed of chemical reactions.

When they stray too close to the line ... you get Intel's 13/14th gen chips that wear out after 1-2 years instead of 10-20 years. Intel calls it "Vmin drift" because that doesn't sound scary, but the actual point is that various wear-out mechanisms push the chip outside of its design envelope - increasing the voltage or lowering the clock speed may get it to run for a while longer, but you're living on borrowed time as the various circuits just stop working right and you get unpredictable instruction mis-execution: https://fgiesen.wordpress.com/2025/05/21/oodle-2-9-14-and-in...

They do degrade physically, but the bigger thing is they stop being competitive quickly. Each year or so we see doubling of GPU speeds for the same amount of power.

If you build a 100MW data center with GPU compute and three years laster a new data center opens with the same cost for GPUs and same electricity cost you do, but can do twice as much compute, you quickly lose business unless the market is just so constrained customers can't afford to be picky. But the moment there's slack in the market you'll see major migrations off of providers that have the same cost but half, or quarter of the same performance.

So when you see someone talking about GPUs fully deprecating in value in 1-3 years this is what they're talking about. Right now it's not a big deal because there's no slack in the market. But once there is, the bottom will drop out.

Transistors do wear out. Not going to elaborate as it is easy to ask GPT

Nothing is stopping them, it's just not worth it: Have a look at e.g. vast.ai's pricing (https://vast.ai/pricing).

The V100 (2017 -> 9 years old) can be rented from $0.02 to $0.37/h (right now I can find a V100 with a Xeon Gold 6140 and 48GB RAM for $0.165/h). Let's assume the guy you rent it to pins it at its 250W TDP and let's ignore the running costs of CPU/RAM/etc... Then you draw 1/4 kwh for that compute hour. The industrial electricity prices in the US vary between 7.5 and 25 ct per kwh (depending on state, time of day, etc...), so at 100% efficiency, assuming nothing ever breaks, and the CPU consumes 0W you earn about 14ct/h.

And remember: V100s hours are sometimes sold at 1/10th the price.

If I pick average conditions you need to start thinking of whether it is worth it to rent them out: Usually it isn't unless you have them anyways and just sell idle capacity.

It's barely worth it to run them in a pure "is it profitable" sense, if we also account for the opportunity cost of taking up a slot in your datacenter it seizes to be worth it really quickly.

Today's data center GPUs are essentially overclocked, and so at limit of how much the chip materials can physically handle, and therefore degrade over time. For example, GH200s operate at 1W/superchip but the actual safe power is somewhere around 650W which will allow them to function for a decade or more. But that leads to around 15% slowdown and that is unacceptable in today's competition. So current GPUs are destined to be depreciating assets.

In future, we might have fixed cost GPUs but not today.

Chips do deteriorate and fail naturally at datacenter scale or in timescales of decades, though not exactly like on financial reports. Leak current increases or electro-migrations occur at junctions or whatever those words mean.

And yeah, it does feel like GPUs will start losing values slower going forward with Moore's Law being dead for a while. It used to be that 3-5 years old GPUs were more useful as space heaters than GPUs, but that's much less of the case today.

GPU do depreciate indeed, but here the depreciating commodity is the token, not the hardware. You sell cheaper token with the same hardware

I assumed the issue was similar to crypto mining, where given finite amounts of space and power it makes sense to always be running the latest and most powerful GPUs instead of keeping older hardware running. There's definitely a secondary market for these GPUs as well.

Your laptop doesn't have a 100% duty cycle. If you ran it like a data center it would indeed wear out much faster.

> There are no moving parts, I dont think memory chips or GPU chips deteriorate naturally

I believe they do, but I too would love to know more details because there are several ways this can happen. Electromigration, package failures, VRAM failures, dielectric breakdown... Hopefully there will be studies soon similar to that old Google paper on HDD failures!

the hardware itself is still useful, but random failures happen every so often, so if you're trying to run a fixed sized fleet then your fleet shrinks when you can't get spares any more

Yes, even if the hardware is untouched. As technology advances, the power cost per compute cycle goes down. A gpu using old tech costs progressively more to operate compared to the newer models. So its value goes down over time = depreciation.

As for duty cycles, the chips are perfectly happy at 100% operation. Cooling and power componants fail, not the chips. But it costs manpower to repair such things and manpower is inconveniant these days. A gpu with any sort of fault just gets dumped.

They'll ban them because (unless run locally or self-hosted) they are just data capture tools for the China.

The “you wouldn’t download a car” meme applies here

They aren't going down, but in the meantime they'll cover their ass by bribing their way into the S&P 500 and then use your 60 year old mother's 401k and teacher's pension to fund their risky capital expenditure.

Today's frontier models will be tomorrows low-end option. I think whatever model you are using today will be less expensive to use a year or two from now.

Yes, you can. There are multiple inference providers out there. The problem is, it’s hard to beat the Chinese providers in cost. And you also have to compete with frontier model providers’ subsidized offerings.

Here's a free startup idea: operate an open-weight model service, and offer "Verified AI Integrity," which signs the input tokens, the seed for the randomness in selecting outputs, and the model ID, proving that the result of the call to AI was completely "organic" and was not interfered with.

Your main audience would be snake oil salesmen trying to prove their AI products are unbiased and not under the thumb of any outside influence. This doesn't address the biases of the model itself, but that's not your business. Your business is selling tokens and security certificates. If you can get the right angel investor, you could maybe have your new standard required for some government applications.

There are plenty of US-based inference providers available, including AWS, that serve Chinese models at competitive prices (vs frontier US models). They also have lots of usage. Not necessarily for coding, but for other enterprise tasks.

It's called AWS. Bedrock is right there. Price or data policy is never the issue. The models themselves are the problem -- most large US companies are not going to touch them.

Source: directly involved in these discussions. You can downvote as much as you'd like but you can't ignore the facts.

Makes me think of how my Claude.md files specifies to use the built in framework code-generators (rails). Those generators are deterministically right every time.

I wonder how often the Agent actually follows the guidance. I do see them follow it when I look. But it doesn't seem so every time.

A lot of the time if you're copying code from one place to another what you actually want to do is abstract it so you can reuse it in both places.

The LLM can easily do this type of stuff, just tell it and it'll happily do it. This is exactly what I mean when I tell people they need to work closer with the AI, tell it how to do things. Don't just tell it what to do and get frustrated when it does it differently than you would.

A good way to achieve this without writing huge prompts is tell it to plan the change first. Just give it some vague low-effort directions. It'll usually get most things right, you tell it what you want different and once you're happy you tell it to go ahead.

AI may get so commoditized for certain use cases that you will not even be able sell inference at a profit. AI might be bundled in with other services, just like cursor bundles in their own AI model for auto complete with their editor. I.e. cameras might have AI for image recognition bundled in etc.

I agree with all of this.

So my question remains the same: How are the players investing 100s of billions in buildout going to hope to make this back? Market capture looks bleak, inference looks like a race to the bottom. End users look like they could be beneficiaries. Where do the big boys go?

> Speculation here but I think openAI/antrhopic api inference is insanely profitable, it just needs more volume to amortize the training costs.

Well, they just rent their hardware, so I'm not so sure. But they'll both be public soon and we should get that breakout in their cost structures, somewhat.

The majority of Deepseek providers on OpenRouter for v4 pro are in the US. Especially interesting is that they are in the same ballpark for pricing.