The AI Data-Center Bust Will Look Like a Boom

When I was fourteen I bought Bitcoin miners with my own money, and watched newer chips kill them on a spreadsheet while Bitcoin was booming.

NVIDIA is arranging half a trillion dollars to finance GPUs, neocloud stocks are ripping, and Etched has hit a ten billion dollar valuation for an ASIC built to take GPUs' steadiest work. OLIX raised $312 million at $3.3 billion eleven days later.

I am bullish, in a way I never was about mining. About half of US adults now use AI chatbots and a quarter use them daily. Nobody halves the token budget, and a better model can create its own market. Custom inference silicon looks to me like one of the better hardware bets of the decade, because the volumes are finally large enough to pay for a chip that does one thing. A quarter of one rich country is still a small paying base for the money being committed against it. None of that saves an individual machine.

On 1 September 2014, Bitmain posted an announcement on the Bitcointalk forum: computing power from its Antminer S2 miners, sold through Hashnest at 0.0016 BTC per GH/s, maintenance fee deducted daily. I was on the other side of that page. I bought the machinery and not the coin: Antminers, steel boxes built to perform one calculation, and Hashnest claims on somebody else's machines. Later, GPU rigs for Ethereum. None of them died of weak demand. They died on a spreadsheet, at five cents a kilowatt-hour, with Bitcoin near its high.

Twelve years later, NVIDIA started talking like a bank. On 10 August 2026 it announced plans with six of the largest names in capital markets, Apollo and BlackRock and Blackstone among them, to mobilise more than $500 billion of outside capital. The agreements were not final. Jensen Huang then said NVIDIA might, project by project, support the value left in the machines for up to 25% of one financing opportunity. So the firm selling the chips is now helping to organise the money that buys them, and residual value is part of the pitch.

Which work sits still#

Serving a token is mostly a memory problem. The machine reads the model's weights and the running conversation out of memory for every token it emits. Much of the accelerator sits waiting while that happens.

The work that has moved so far is boring. Meta says its MTIA chips serve recommendation models while unsupported models remain on GPUs. Amazon reported in April that it had landed more than 2.1 million AI chips in twelve months, more than half Trainium, alongside more than one million announced NVIDIA GPUs from 2026. It buys both, which is what you would expect if the jobs are different.

My bet through 2030 is narrower than "custom chips win". The largest platforms move stable, repeated, high-volume inference onto chips they control. GPUs keep the models that change, the software those chips do not support, the demand spikes, the customers who need broad compatibility. Move enough predictable queues off a fleet and the machines carry on running perfectly well while the spreadsheet that justified buying them stops working.

Google is the obvious objection, and the best one anybody has put to me. It has served its own steady inference on chips it designed for about a decade. NVIDIA's data-center business compounded through the whole period, with Google buying GPUs the entire way. If migration alone were enough, my bust would have arrived years ago. So the call is about who owns the fleet that loses the work, and on what terms: borrowed money, against a six-year book.

I am wrong by 2030 if custom chips stay too painful to program, if model changes eat the savings, or if old GPUs keep named workloads and steady cash yields through a major hardware transition. Any one would be enough, and growing demand settles none of them.

Model change is the one that worries me. The calculation my Antminers raced to perform never changed, which is the only reason a chip welded to it could win. Transformers are not fixed. Mixtures of experts, longer context and models that think for longer before answering all make the work less regular. Weld too early and the chip arrives fitted to last year's assumption. None of this needs a view on whether AGI is coming: if the models keep compounding, the predictable work migrates sooner, and if they stall, the debt still outlives the fleet.

The machine dies on a spreadsheet#

Energy per hash fell about 984-fold between two sparse endpoints, a 2013 Avalon at 9,351 joules per terahash and 9.5 for Bitmain's 2025 S23 Hyd. That is the curve an older machine had to keep clearing its power bill against.

Using April 2026's average Bitcoin network revenue and electricity at five cents per kilowatt-hour, a 29.5 J/TH Antminer S19 Pro failed to cover electricity alone. A 13.5 J/TH S21 XP still cleared it. Both worked. Only one paid its power bill, and cooling, labour, financing and the price of the machine were still waiting outside the calculation.

Miners have a name for a machine on the wrong side of its line: a space heater. The name is unfair, and that unfairness is the best objection to everything I argue after it. A rig underwater at one power price clears at another. Operators undervolt them and move them to power nobody else wants, and rigs written off at five cents were still hashing years later. Cheap electricity buys an old machine more time. It does not find it a job.

Every miner knew this. MARA wrote the mechanism into its own annual report:

As new and existing miners deploy additional hash rate, the global network hash rate will continue to increase, meaning a miner's share of the global network hash rate (and therefore its chance of earning bitcoin rewards) will decline if it fails to deploy additional hash rate at pace with the industry.

— MARA, Form 10-K, 2023

The resale market moved before the machines wore out. During 2022 Hashrate Index recorded three ASIC price indexes falling between 82% and 87%. Bitcoin carried on. The miners carried on. The boxes were worth a fraction of what they had cost.

The chip wins, the vendor usually does not#

Being right about a chip is not the same as owning the company that sells it. Mining ran that experiment for a decade. Demand for mining chips beat every forecast: computing power on the Bitcoin network rose about 351 times between January 2017 and April 2026. Knowing that in 2013, you would have bought the vendors.

You would have lost almost all of it. Revenue per unit of that computing power fell 95% across the same months, out of the same file. Only the second line pays for anything.

The buyers went first, diluted by every machine that shipped after theirs. Then the vendors went. KnCMiner, a Swedish maker of Bitcoin ASICs, raised $32 million from investors including Accel and Creandum, shipped real machines to real customers, and declared bankruptcy in May 2016. Its chips worked. Its chief executive blamed the coming halving: "Effectively our cost of coin...will be over the market price".

Three clocks on one machine#

Microsoft reported $41 billion of company-wide capital spending in its fourth quarter, roughly two thirds of it in short-lived assets. NVIDIA reported $75.246 billion of Data Center revenue in its latest quarter. Broadcom booked $10.8 billion of AI semiconductor revenue and guided the next quarter to $16 billion.

Amazon extended server lives from five years to six in 2024, then shortened a subset to five a year later because AI and machine-learning technology was changing faster. The reversal added $1.4 billion to 2025 depreciation and amortisation expense. CoreWeave spreads technology-equipment cost across six years while its weighted-average committed customer-contract term was about five years at the end of 2025.

The reversal is a footnote in the 10-K:

Effective January 1, 2024, we changed our estimate of the useful lives for our servers from five to six years, and effective January 1, 2025, we changed our estimate of the useful lives of a subset of our servers and networking equipment from six to five years.

— Amazon, Form 10-K, 2025

The filings do not show a cemetery of formally impaired GPUs. I am not claiming one exists. Only that the accounting life, contract life and earning life of one machine already run on three different clocks. There is no line item for space heaters.

The pot is where the analogy breaks, and I would rather break it myself. Bitcoin pays its machines out of a pot designed to shrink, so the divergence between rising capacity and falling revenue per machine is close to an accounting identity. AI has no such ceiling: more compute really does produce more finished work. What I would still defend is narrower. A better machine arriving to do the same job at a lower cost per unit of work killed my Antminers, and that happened in rising markets about as often as falling ones.

Who gets paid when the chip changes#

One rule sorts this industry: prefer the business that gets paid when the chip changes, and distrust the business that needs the chip not to change before the debt is repaid. Bitmain passed that test and its customers did not. The merchant vendor is sorted above, which leaves three positions.

The chip seller has the safest seat, because it is also selling the system around the chip. NVIDIA is working to make its network useful around accelerators other companies design, while Microsoft says Maia 200 uses its own transport over standard Ethernet, which reads to me like a polite way of routing around it.

The borrower who owns the fleet is where the analogy bites hardest. CoreWeave is the cleanest public expression. At the end of June it disclosed $35.6 billion of debt principal and $16.3 billion of recognised operating-lease liabilities, against $14.1 billion spent on equipment in six months. The defence is serious: $103.7 billion of work still owed under contracts, with 98% of quarterly revenue tied to customer commitments. The other side is that three customers supplied 72% of Q2 revenue and quarterly interest expense was $640 million. A backlog does not pay interest. Contract cash has to cover interest, leases, replacement chips and the cost of finding the next job. That arithmetic belongs to every leveraged fleet in the industry, listed or not.

It is the first item on CoreWeave's own list of risks:

Our substantial indebtedness could materially adversely affect our financial condition, our ability to raise additional capital to fund our operations, our ability to react to changes in the economy or our industry, and could divert our cash flow from operations for debt payments.

— CoreWeave, Form 10-Q, 30 June 2026

In March CoreWeave closed an $8.5 billion facility at investment grade, SOFR plus 2.25, against GPUs wrapped in a customer contract. In August it closed a $2.6 billion facility at SOFR plus 5.50, below investment grade, a roughly five-year loan against contracts averaging about three years, with no residual-value guarantee anywhere in the filed agreement. The 325 basis points between those two loans is a first public price for the years a GPU must earn after its contract ends.

The landlord is a different animal and the market keeps pricing it as the same one. Applied Digital's core leases sell the powered building rather than the accelerator inside. A mine leaves behind a power connection and not an inference customer, which is why this is the position I would rather hold through a hardware transition. The condition is that only 100 of 1,410 contracted megawatts was operating and earning revenue at year-end, so those contracts still have to become buildings.

What happens next#

Predictions, so this essay can be graded. Footnotes arrive quarterly on a ledger that logs the prints cutting against me too.

The six-year GPU book will not survive the decade. Amazon's cut from six back to five was the first crack; before 2030 a hyperscaler or major neocloud discloses four or fewer for accelerators, and the sell side will call it prudence when it is a confession. Miners tried five and two and met at three.

AI will get its hashprice. The H100 rental series fell 57% in under three years, and the sceptic's answer rides in the same series: one-year contract rates up 38% from their October low, on-demand capacity sold out since February. None of it is yet a number a lender marks a fleet against, which is what made the difficulty adjustment lethal. One of these series hardens into that number before 2030, and then the private cliff becomes a public one.

The best answer to me is not mine. It came from CoreWeave's chief executive, talking about a card two architectures old:

The fact that we have been able to go ahead and sell a GPU who's architecture was from 2020, in a contract that was fully priced out to 2029, really provides some insight into what the future is going to look like as this infrastructure comes off contract.

— CoreWeave, Q2 2026 earnings call

Most neoclouds will not die as clouds. The survivors will quietly become power and shell companies, the way the surviving miners did, because the durable asset underneath was always the grid connection and the building. The rest get absorbed for their contracts and their megawatts, with the GPUs going to the secondary market.

The bill arrives before the boom ends#

At fourteen I thought I was buying computing power. What I had bought was a temporary spread between mining revenue and electricity, wrapped in a machine. Turner painted the asset class in 1839: a working ship, nothing wrong with her, under tow behind the machine that took her job.

The AI data-center boom is a repricing event dressed as a building programme: hundreds of billions of dollars a year underwritten on the belief that a general-purpose chip keeps its best job for five or six accounting years, when everything mining taught me says the best job leaves first, quietly, while the buildings are still full. Somebody ends up holding the one asset this industry has never learned to price: a space heater with a six-year book.

Before I underwrite another GPU rack, I want one answer:

Which named customer and workload will still pay for this machine if custom chips take more of the predictable work?

If the answer is "AI demand", I have seen this asset before. It was a mining rig with better branding.

If I am wrong, the ledger will say so before I do.

Braindump by Josef Chen

Research cutoff: 16 August 2026. This is a research view, not investment advice.