You Are Financing the AI Revolution. Nobody Asked You.

There is a man you should know. His name is Michael Burry. You may have seen him in the movie The Big Short — played by Christian Bale, wearing a single glass eye and a T-shirt, sitting alone in a glass-walled office, headbanging to heavy metal while his colleagues thought he had gone completely mad.He was not mad. He had simply read something nobody else had bothered to read: the fine print inside American mortgage bonds. He saw that thousands of shaky home loans had been bundled up, given a shiny credit rating, and sold to pension funds and insurance companies as "safe" investments. When the housing market cracked in 2008, those bonds blew up — and the holders, mostly ordinary people, paid the price.Burry made $750 million by betting against them.Sixteen years later, he is saying something similar about AI.

Part One: What Burry Actually Said:

This week, Burry used the word "fugazi" to describe Nvidia's AI financing structure. Fugazi is American slang — it means something that looks real but is hollow at its core. He was not calling Nvidia's chips fake. The chips are very real. He was calling the financial pipeline around them misleading.

Here is the structure he traced, in plain language.

Nvidia sold approximately $5.4 billion worth of its most powerful AI chips — called GB200s — to a company called Valor. Now, Valor is not a famous tech company you have heard of. It is a Special Purpose Vehicle, or SPV. Think of it as an empty shell company created specifically for this one transaction. It has no other business. Its only job is to hold these chips.

Valor then borrowed most of the money it needed — around $3.5 billion — from Apollo, a very large American financial firm. Nvidia itself put in the remaining $1.9 billion as an equity stake. So Valor now owns the chips, financed partly by Apollo's debt and partly by Nvidia's own money.

Valor then leased those chips to xAI — Elon Musk's artificial intelligence company — for five years on what is called a triple-net lease. xAI pays rent every month. xAI does not own the chips. xAI does not carry any debt on its balance sheet. The debt is Valor's problem.

This is where it gets interesting.

Apollo is not just a lending firm. It also owns an insurance company called Athene. Athene's primary business is selling retirement annuities to ordinary Americans. You give Athene a lump sum when you retire — say, $200,000 — and Athene promises to pay you a fixed monthly amount for the rest of your life.

To fund those monthly payments, Athene invests your money. And some of those investments are now, through the chain above, backed by xAI's ability to keep paying rent on Nvidia chips for the next five years.

The American retiree reading their monthly statement has no idea. They just see a number.

Every single link in this chain is legal. Every part is publicly disclosed. Nobody is hiding anything. Burry's point is not that someone broke the rules. His point is that the rules allow something very uncomfortable to happen — and nobody is asking whether the people at the end of the chain understand what they are holding.

Part Two: The Special Purpose Vehicle — What It Is and Why It Matters

Before going further, it is worth understanding the SPV properly, because this structure is the heart of the story.

A Special Purpose Vehicle is a company created for one specific financial purpose. It is like setting up a separate entity just to own one asset — in this case, AI chips. The benefit for the creator is clean and deliberate: if the SPV fails, it does not drag the parent company down with it. The risk is ring-fenced inside the shell.

This is not a new trick. SPVs were central to the 2008 housing crisis. Thousands of individual home mortgages — many of them given to people who could not really afford them — were bundled together inside SPVs and sold as bonds. Each bond had a credit rating. The rating said "safe." The underlying mortgages were not safe. When the housing market fell, the bonds collapsed, but the banks that originally created the SPVs had already been paid and had moved on.

The structure here is not identical. But the principle is the same: create a vehicle, put the asset inside it, sell the risk downstream, and let the original creator walk away clean.

Nvidia ships the chips and books full revenue on day one. After that, the risk is Valor's. Valor's risk is backed by Apollo's loan. Apollo's exposure sits on Athene's books. Athene's health backs the retirement payments of people who signed up for a guaranteed monthly income.

The chips moved forward. The money moved forward. The risk moved forward. And at each step, the new holder understands it slightly less than the previous one.

Part Three: The Four Clocks Problem

Imagine you take one photograph of a river. Four different people look at that photograph — but each person is looking at it through a different time-delay filter. The first person sees the river as it is right now. The second sees it as it was three months ago. The third sees it as it was a year ago. The fourth sees it as it was whenever someone last wrote a formal report about it.

They are all looking at the same river. But they are seeing four different rivers. If the river dries up tomorrow, the first person finds out immediately. The fourth person might not know for two years.

This is exactly how the Valor–Apollo–Athene–Retiree chain works.

Nvidia's clock runs on the trade. Revenue is verified the moment the chips ship. Their exposure ends there. They have been paid.

Apollo's clock runs on the lease coupon. They check the books every quarter when xAI pays its rent. As long as the coupon arrives, everything looks fine on paper.

Athene's clock runs on the model, the credit rating, and the Bermuda solvency framework. They are audited roughly once a year. A ratings agency looks at the portfolio, confirms the structure is sound, and signs off.

The retiree's clock runs on a monthly statement. They open an envelope or log into an app. They see a number. That number reflects what the model says the portfolio is worth — not what a live market would price it at today.

Same machine. Four clocks. Four different prices being printed simultaneously for what is, at the bottom, one exposure to one AI buildout cycle.

When something goes wrong in this chain, Nvidia already knew. Apollo figures it out within a quarter. Athene discovers it on the next annual audit. The retiree finds out last. By then, the damage is already done and cannot be undone.

Part Four: Jensen Huang's Half of the Truth

At Computex 2026 in Taipei this week, Jensen Huang said that fears of AI job loss are "complete nonsense." He showed data from GitHub. Software commits — the number of code updates made by developers globally — went from 300 million in 2023 to 500 million in 2025 and nearly 1.4 billion in just the first few months of 2026. AI tools have made every developer dramatically more productive, so companies are hiring more of them, not fewer.

His logic is clean and his data is real. If one engineer now produces three times the output, the return on hiring that engineer has tripled. So you hire more. The math works.

Jensen is not lying. Senior engineers are being augmented. Companies building AI products are growing their headcounts. The total number of software jobs is not collapsing in aggregate.

But there is something he is not saying.

The data tells a different story at the bottom of the skills ladder. Workers aged 22 to 25 in AI-exposed occupations have seen employment fall 13 to 16 percent relative to trend over the last three years. Older workers in the same roles? Essentially unchanged.

AI is not destroying the job market. It is repricing the bottom of it. The freshers. The junior analysts. The entry-level coders. The work they used to do as their learning curve — the first two or three years of building professional judgment through repetition and low-stakes tasks — is now being done by an AI model for a fraction of the cost.

Conclusion: Both Men Are Right. They Are Watching Different Parts of the Same Machine.

Jensen sees the surface clearly. The platform is real. The productivity is real. Over a long enough arc — a decade, two decades — work will expand and the economy will absorb the displacement. He is almost certainly correct about the long term.

Burry sees the plumbing clearly. The risk of a buildout cycle does not disappear inside an SPV. It moves — through each link of the financial chain — to the holder with the slowest verification clock and the least ability to act when they finally find out.

The chips are real. The revenue is real. The displacement of entry-level workers is real. And the credit risk is being routed, quarter by quarter, into the books of people who were never asked whether they wanted to finance the buildout.

Financial structures always look stable until the moment they do not. The four clocks always resynchronize eventually. The only question is who is standing in the gap when they do.

Right now, that person is holding a monthly statement. Reading a number. Trusting the sytem.

As investors — and as citizens — we should at least know their names.

This article is for financial education only. Not investment advice