AI's Trillion-Dollar Black Hole: Money is Being Burned, But Besides Two Money-Losing Companies, Who Is Actually Paying?

Original Author: Ed Zitron

Original Translation: TechFlow

Introduction: The Bank for International Settlements (BIS) annual report reveals a truth that tech giants have deliberately ignored: trillion-dollar AI capital expenditures have already surpassed these companies' cash flows and profitability, and the ultimate destination of this money is simply to sustain OpenAI and Anthropic, two model labs operating at massive losses. If Microsoft, Google, and Amazon decide to stop their quarterly $30 billion GPU purchases, the entire supply chain will collapse.

This month, I published a two-part in-depth series analyzing the "bubble-within-a-bubble" structure of the AI bubble – from the unsustainable and reckless growth of semiconductor companies to the personality cult surrounding Sam Altman and Dario Amodei. On Friday, I'll be publishing the long-awaited "The SoftBank Hater's Guide," and you won't want to miss it.

On Sunday, the Bank for International Settlements (BIS) released its annual report, saying a lot of what I've been saying:

In the short term, the ongoing AI investment boom raises questions about the sustainability of the current economic expansion. The five major hyperscalers plan to spend over $1 trillion on AI-related capital expenditures between 2025 and 2026. These commitments exceed these companies' profits and free cash flow, leading some to issue debt to raise additional funds.

It's reassuring to see the central bank of central banks articulating what I've been saying for the past few years. This part, however, both validates my stance and makes me feel terrible for the whole world:

Disappointing returns could trigger a sudden pullback in financing and transform the capex boom into a prolonged investment slump, with potential knock-on effects on financial conditions... If hyperscalers slow or halt their aggressive capex deployment pace, many borrowers in the supply chain may struggle to replace lost revenue and repay their debts.

Nonsense. Last April, I wrote an article titled "AI is a Systemic Risk to the Tech Industry," outlining how the failure of one model lab, OpenAI, would have earthquake-level effects on its supply chain, delivering blow after blow to Nvidia, Oracle, Microsoft, and the various new cloud providers (most notably CoreWeave) supplying them with compute power.

Since then, OpenAI's sticky tentacles have reached further into the tech industry, signing agreements with companies like Google, Amazon, Cerebras, and Broadcom, while also accepting more investments, including a massive commitment from SoftBank. SoftBank can only fulfill these commitments by selling precious shares of companies like ARM and Nvidia and by taking on debt.

The concept of systemic risk has never really left my work; I've spent much of the past year thinking about it – hence my writing has examined the potential consequences of an AI spending pullback on the financing side of the industry, particularly private credit, and the semiconductor sector.

The BIS isn't concerned about a revenue crash – which would happen if hyperscalers "slow or halt their aggressive capex deployment pace," as they fear – but about a revenue crash combined with the inability of borrowers in the AI supply chain to repay their growing debt burdens.

Again, this is something I've sounded the alarm about many times. CoreWeave has been a favorite of this newsletter; in March 2025, I published "CoreWeave is a Ticking Time Bomb," focusing on the company's overwhelming toxic debt pile and its reliance on OpenAI as a customer.

On a larger scale, we have Oracle – a company I covered extensively in the "The Oracle Hater's Guide" newsletter.

Unlike new cloud providers like CoreWeave, Oracle is an older company that, for most of its existence, sold database and ERP software to some of the world's largest companies and public sector institutions. Oracle pivoted to offering AI compute power when its core business lines began to stagnate, and due to its massive size, it could raise staggering amounts of debt.

As I've pointed out before, Oracle was a heavily indebted company even before the AI bubble. Coincidentally, because of its dalliance with OpenAI, Larry Ellison felt the need to crank the debt dial up to eleven.

Oracle's spending has pushed its free cash flow into negative territory – negative $23.7 billion as of the end of fiscal year 2026 – and it had $129.5 billion in outstanding debt as of the end of May. This doesn't include its various lease commitments, totaling nearly $38 billion, nor an additional $260 billion in lease commitments signed but not yet actually commenced.

All of this is to say that Oracle has taken on massive debt for the benefit of one company, OpenAI, and if that company can't pay its bills, Oracle is finished. Oracle's existence – and Larry Ellison's personal wealth – depends on OpenAI honoring its promise to spend $300 billion on compute power.

This is both the most obvious and the least discussed part of the AI bubble – the over $1 trillion in capex from hyperscalers is fueling a massive semiconductor boom, which is based, at best, on the highly improbable assumption that large language models will morph into something completely different.

If Microsoft, Google, Amazon, and Meta decided to stop spending $30 billion or more per quarter on GPUs, RAM, storage, and data center construction, it would tear a hole in the side of what people assume is a permanent super-cycle.

I need to emphasize how foolish it is to think that the so-called semiconductor boom is anything other than a fleeting opportunity to fill your boots before a global stock market catastrophe. A catastrophe so severe it would make Futurum Group want to off itself.

The hyperscalers – whose capex will exceed their cash flow by Q3 2026 – are getting such poor returns on their AI investments that none of them will actually disclose revenue beyond vague "annualized revenue" figures, implying that all these investments are based on the idea that something completely different will happen in the future.

This future must generate at least $2 trillion in entirely new revenue for them by 2030, because if it doesn't, practically all capital expenditure will have been used to prop up Anthropic, OpenAI, and whatever Meta is doing with its chatbot.

There is no compelling or rational argument to continue capex, at least not one that doesn't implicitly accept that most current spending is wasteful, serving only to inflate stock prices and incubate two different large, loss-making AI labs. Those millions of H100, B200, and B300 GPUs won't usher in a digital god; they won't create recursive self-improvement; they won't become the fulcrum for adding $600 billion or more in entirely new revenue to current services. The only revenue they generate comes from compute spending by Anthropic and OpenAI, which I estimate accounts for 20% or more of the cloud revenue for Google, Amazon, and Microsoft.

I must also be clear that the costs for these companies extend far beyond equity investments. While Microsoft invested $13 billion in OpenAI, Microsoft executive Michael Wetter revealed in the Musk v. Altman case that the partnership has cost over $100 billion, implying that just OpenAI's infrastructure costs are at least $87 billion. I imagine Amazon and Google must spend similar amounts to handle Anthropic's equally voracious compute needs, especially considering Amazon's cost of $11 billion and more for the Anthropic-dedicated Rainier project data center.

This is the severely under-discussed part of the AI bubble. Anthropic and OpenAI have collectively raised less than $300 billion since 2019, but I estimate their true cost is at least $500 billion, considering the hyperscaler capex investments necessary for their existence. This doesn't even account for the $340 billion or more Oracle is spending to build the 7.1GW "Stargate" data center for OpenAI. These are not startups; they are subsidiaries of big tech companies, existing as independent divisions solely to prop up equity positions and hide the truth: AI capex is a complete waste of money, even if you factor in two obese spendthrifts losing tens of billions of dollars annually.

As I reported two weeks ago, OpenAI spent $17.2 billion on Microsoft Azure in 2025, a year in which it lost $20.9 billion on $13.04 billion in revenue. Even if that were profit (it isn't), it's $4.2 billion less than Microsoft's capex in Q1 2025 alone.

Beyond OpenAI, Microsoft arguably has no AI business. While it boasted of $37 billion in AI annualized revenue in April (meaning a non-specific month multiplied by 12), that's only about $3.08 billion per month, or less than one-tenth of its $31.9 billion capex for that quarter. Worse, Microsoft revealed this number was "up 12% year-over-year," implying its AI annualized revenue in Q3 fiscal 2025 was $16.59 billion, or about $1.38 billion per month.

Yet my November report on OpenAI's inference spending showed it spent $2.947 billion in Q3 fiscal 2025, annualizing to about $11.7 billion, meaning that at least in that quarter, OpenAI likely accounted for about 70% of Microsoft's AI revenue. I'd be surprised if there was a dramatic change over the year, given OpenAI's inference spending was $3.648 billion in Q1 fiscal 2026.

All of this is to say, the only real outcome of all this capital expenditure seems to be propping up two deeply unprofitable companies, Anthropic and OpenAI, and then recouping a small fraction in the form of revenue, which is only made possible through hundreds of billions in venture capital subsidies.

Now OpenAI and Anthropic account for 50% or more of the hyperscalers' remaining performance obligations, approximately $748 billion.

Beyond the mistaken belief that OpenAI or Anthropic can actually afford to pay without Google, Amazon, or Microsoft giving them money, there is simply no logical or rational reason to further invest capex in AI. The hyperscalers have no meaningful AI revenue in any form other than their own pseudo-startup investments. That A) they continue to invest and B) the market, analysts, and journalists act like everything is fine is both absurd and irrational.

Side note: I'm not discussing Meta because Meta has no AI story. Mark Zuckerberg is wasting every ounce of its capex, beyond anything it might get from reselling capacity to others – but don't worry, he thinks (this is a quote!) Meta has a use for compute! No, sorry, those GPUs aren't driving meaningful ad revenue growth, and I've talked about that in the past.

The record sales of Nvidia, Micron, SanDisk, SK Hynix, and Samsung are a direct result of a purely speculative asset bubble, driven by the reckless and directionless capex of some of the world's largest and richest companies.

Anyone investing in data centers is building speculative capacity for a demand that doesn't exist beyond Anthropic and OpenAI. If such demand existed, the new AI data center cloud company CoreWeave would have healthy and diversified revenue streams, instead of 65% of its revenue coming from Microsoft (for OpenAI) and Nvidia, with the rest from Google (for OpenAI), Anthropic, Meta, and of course, OpenAI itself. There are simply no other large-scale consumers of AI compute power, and the only reason we haven't hit that harsh reality yet is that data centers take 18-34 months to complete.

Even so, I can find almost no evidence of anyone besides OpenAI, Anthropic, and the hyperscalers having the demand or capital required to justify data center construction.

I really need to emphasize this.

If we assume Nvidia CEO Jensen Huang's prediction of $1 trillion in Blackwell and Vera Rubin sales comes true, that would be approximately 40GW of data center capacity, with an IT load of about 30GW. If we generously assume data centers generate about $12 per megawatt in revenue, that would create about $435 billion in annual compute demand by 2030.

Let's be very clear about one thing: the only companies that can currently afford to spend money on compute are either hyperscalers or companies subsidized by hyperscalers. Even so, beyond OpenAI's projected $50 billion in compute spending in 2026 and my estimate of a similar amount for Anthropic, there seems to be no more than a few tens of billions in demand, if that. Otherwise, CoreWeave, IREN, Nebius, Cipher Mining, and other new cloud providers would have hundreds of billions in remaining performance obligations, not ones that only expand under hyperscaler backing or Meta's Zuckerberg-esque AI psychosis.

Let me put it more simply: those hundreds of billions in data centers are being built for no one. The only companies that can "afford" to pay even a fraction of the compute costs are unprofitable AI companies propped up by hyperscalers.

While this might read like a radical stance, I think looking at the current state of affairs and saying "screw it, I think hyperscalers should spend a trillion dollars next year" is far more radical.

Driven by a fantasy mindset desperate to avoid thinking that tech has no high-growth ideas, there is no rational reason to do so.

Aside from creating OpenAI and Anthropic, the current capital expenditure is almost entirely wasteful. Microsoft 365 Copilot sucks. GitHub Copilot sucks. Google AI Overviews sucks. Google Gemini is a follower LLM and therefore also sucks. Meta's LLMs are very dangerous. Amazon Rufus sucks, and Amazon should be investigated by the SEC for suggesting it drove $10 billion in "annualized revenue" in Q3 2025, because it absolutely did not. Alexa+ sucks. Everything sucks, and it would be just as bad if big tech had spent only a quarter of the capex.

These products are almost universally disliked, generate almost no revenue, and even in the modestly successful case of GitHub Copilot (about $1.08 billion annualized by the end of last year), that's only because users' compute was heavily subsidized, leading Microsoft to move users to token-based billing, angering customers used to burning thousands of dollars in tokens for a flat $39 a month.

Yes, All This Money Could Be Wrong

Sundar Pichai, Andy Jassy, Satya Nadella, and Mark Zuckerberg are losers. They may have billions of dollars, they may run giant tech companies, but they are losers, selling a technology doomed to fail, based on unreliable, inefficient, and overly expensive technology unsuitable for the reliable, deterministic, "set it and forget it" nature people actually associate with AI.

The four big losers are the only reason anyone takes these big, losing models seriously, which is a sign that the tech industry and our economy are also being driven by losers. Every bit of "progress" we see in LLMs comes from forcing square pegs into round holes – billions in training costs, hundreds of billions in capex, endless tools, scripts, wrappers, and layers trying to squeeze out anything resembling the promised autonomy.

All the king's horses and all the king's men are pouring every dollar and ounce of brainpower into trying to make LLMs into something they are not. As a society, we are expected to coddle these things, act like they are great, and give them credit for things that haven't happened. I reject the premise that an LLM's ability to generate code or replicate open-source software is evidence that they will one day become powerful autonomous tools. I consider anyone who extrapolates to that point to be intellectually bankrupt, deeply cynical, or gullible enough to click every email claiming their PayPal account has been compromised.

I assure you, the money could be wrong! Hyperscalers can indeed spend a trillion dollars on something that doesn't do what it's supposed to, because these companies are perfectly happy to mislead you, to quote Nik Suresh:

A large part of the economy is driven by people who are, simply put, very suggestible. That is, it's very, very easy to get them excited and willing to spend money.

Why is everyone investing in data centers? Because the hyperscalers are doing it! Why are Micron and memory companies selling so much memory? Because A) GPUs use a lot of HBM, B) that HBM consumes three times the wafer space of regular DRAM, leaving less room for other, cheaper, lower-margin memory types, and C) because the servers for these AI GPUs are also packed with memory!

These data centers are not being built because creditors have any "insight" into the massive AI compute needs that generative AI tools will require. They saw the "success" of ChatGPT and Claude (two heavily subsidized products) and thought that because Anthropic and OpenAI need tons of compute, everyone will need tons of compute. And because banks and private credit are desperate for ways to invest, everyone is so excited that it's