Before we begin, I want to run a small experiment. Imagine you sell your car for $40,000. Good price, clean title, money in hand. Except the buyer is a company, you put $14,000 into the week before the sale of the car. The buyer does not drive. The buyer exists for only one purpose: to own your car.

The car itself sits in your neighbor's driveway, and your neighbor drives it every day under a lease that makes him cover the gas, the insurance, and the oil changes on a car he does not own. And the money the company used to buy your car? Mostly borrowed. That loan was packaged into a security and sold on, until it landed inside the insurance company that holds your mother's retirement annuity.

Now answer one question. Did you really sell your car? You booked the $40,000. The title moved. Every paper is legal, and every signature is real. But your own money helped fund the purchase; your neighbor is the only one who drives it, and if he ever stops paying, the loss falls on someone who has never seen the car. Hold that picture at whatever size you like. This week, the size is $5.4 billion.

Scene 1: The Sale.

On January 7, 2026, Apollo published the sentence that matters. A fund managed by Valor Equity Partners would acquire and lease $5.4 billion of data center compute infrastructure, including NVIDIA GB200 GPUs, to a subsidiary of xAI. Apollo-managed funds and affiliates would lead a $3.5 billion capital solution. The lease would be triple net. The cluster would train Grok. The chips would run inside Elon Musk's AI machine. Reports put the count above 100,000 GPUs. The most advanced silicon on earth, sold in a single block.

Scene 2: The Middle.

The part that turned a chip deal into a Dead Drop was not the lease. It was the middle. The buyer was not xAI. The buyer was Valor Compute Infrastructure, a vehicle built for one job: to own compute assets. And NVIDIA was not only the vendor. Apollo's own release says NVIDIA invested in the vehicle as an anchor limited partner, a stake secondary reporting puts at $1.9 billion.

Nobody hid any of it. Every piece sat in plain sight, in a press release written for people who would not read past the headline.

Scene 3: The Map.

On May 31, Michael Burry drew the map and gave it a name. He published a diagram on his Cassandra Unchained Substack titled "The Retiree/Apollo/Nvidia/Bermuda/AMAPS/xAI Pipeline," and over it he wrote one line: "It is all Fugazi. How to make tens of $billions worth of $NVDA GPUs disappear from balance sheets in 8-12 byzantine steps."

His claim was not that the structure was hidden. His claim was that it had been engineered so each participant could show the market the cleanest possible version of the same deal.

That is the part worth staring at. The chip did not disappear. It changed costumes. In the factory, it was inventory. In the press release, it was AI infrastructure. In the lease, it was productive capacity. In the capital stack, it was collateral. By the time the risk reached the bottom of the structure, the server rack had been translated into something a retirement portfolio could hold without anyone asking what was inside.

They told you the future was being built. They did not tell you whose savings were holding it up.

AI is moving fast and most people are falling behind. 

The Rundown AI keeps you ahead of the curve. 

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.

Operatives, last week the file was the meter. AI access had become inventory for criminals. Stolen keys, stolen tokens, stolen accounts, resold by the hour until a three-person team woke up to an $82,000 bill. The machine did not know it had been robbed. The key was valid. The endpoint answered. The meter ran. This week the file moves from the stolen meter to the legal one. No credential theft. No breach. No hoodie. Just lawyers, bankers, lease terms, and press releases, moving the most valuable chips on earth through a structure almost nobody outside finance can read on the first pass.

Valor Compute Infrastructure buys $5.4 billion of NVIDIA GB200 compute and leases it to xAI under a five-year triple-net lease. Apollo leads $3.5 billion of financing. NVIDIA invests in the same vehicle as an anchor limited partner. Every piece is publicly disclosed. That is what makes it useful. The trick is not secrecy. The trick is translation.

Michael Burry calls it fugazi. Strip away the theatrical word, and his case rests on three plain worries. The money is round-tripped: the seller helps fund the buyer, then books the purchase as clean revenue. The basket holds one egg: the vehicle owns one kind of chip, leased to one tenant. And the collateral ages fast: GPUs depreciate like fish, not like farmland, and the debt does not depreciate with them. Same architecture as always. A promise sized for the public. A mechanism sized for the house. This time the house is not a court, a data broker, or a cloud console. It is a capital stack. And the floor it stands on is somebody's retirement.

Start with the clean version:

Now read it as a ledger.

NVIDIA records product revenue when control transfers to the customer. NVIDIA's own fiscal 2026 annual report says that is how product revenue works. The customer in this structure is the compute vehicle, not the operating AI company burning the tokens. The machine can sit in xAI's data center. The legal owner can be Valor's vehicle. The revenue can sit at NVIDIA. Those are not contradictions. They are the design.

The triple-net lease is the hinge. In real estate, a triple-net lease makes the tenant responsible for the taxes, the insurance, and the upkeep. Here it makes xAI responsible for the rent and the running costs of the chips it does not own. That turns a rack of silicon into an asset with contracted payments attached. And once payments attach to an asset, finance can package them. Apollo calls the transaction downside-protected and asset-based. Valor calls it quarterly distributions and upside through ownership. Those are not scandal words. They are finance words. They are also the words that reveal the transformation. The AI future has become an income product.

This is where the retiree enters Burry's map. Apollo owns Athene, a retirement-services business that sells annuities to ordinary Americans and manages the insurance balance sheet behind them. Burry's diagram routes the Valor debt through Apollo's asset-backed finance machinery and into that insurance stack, and then further, offshore. His figures, and at this point they are his figures, describe Athene ceding hundreds of billions of dollars of assets to a Bermuda reinsurance affiliate, where leverage runs high and where, he says, more than a third of the assets are Level 3. Level 3 is the accounting label for assets with no market price. You do not mark them to market. You mark them to model. You mark them, in the end, to hope.

Be careful here, because the file demands it. We do not have proof that a specific annuity holder's premium bought a specific slice of Valor's debt. Burry's claim is about routing, not receipts: that the private-credit machine funding xAI's compute and the balance sheet standing behind retirement annuities are parts of the same machine, run by the same house. The defense is obvious and not frivolous. Apollo will say the deal is collateralized, professionally underwritten, and structured to protect the downside. Analysts have already called Burry's framing sensationalized, standard asset-backed lending wearing a scary costume. NVIDIA will say demand is real and the chips are real. Valor will say investors own hard assets with contracted lease economics. All of that can be true.

But NVIDIA's own filings show the ground shifting under the simple story. A year ago, NVIDIA's annual report said the company had been asked to offer financing arrangements to support customers' data-center buildouts and had not entered into any. By the quarterly filing this May, the language had changed. A billion dollars in infrastructure-fund investments accounted for under the equity method. A maximum loss exposure of $2.3 billion on those positions. Total investment commitments of $27 billion. In one quarter, $18.6 billion of investments in private companies and infrastructure funds, including AI model makers that may turn around and buy or rent NVIDIA compute. Guarantees on partners' facility leases with gross exposure of $3.5 billion. The vendor is no longer just a vendor. The vendor is becoming a capital participant in the ecosystem that buys its product.

The Dead Drop question is what happens when all of those true statements depend on the same assumption: that AI demand grows fast enough, long enough, and profitably enough to keep the lease payments looking boring. Because every floor of this structure leans on the floor below it, and the bottom floor is an insurance promise to someone who has never heard the word Valor.

A bubble does not require fake assets. Sometimes the assets are real, the contracts are real, the lawyers are real, and the illusion is the belief that a real asset cannot be overpriced when everyone needs it at once.

Burry's second warning, and why the IPO calendar is part of the same file.

Burry did not stop at the pipeline. In early May, he wrote that the market feels like the last months of the 1999-2000 bubble. Absolutely non-stop AI, he said. Nobody is talking about anything else all day. He has been posting the rhymes ever since. The S&P 500's 10-year earnings multiple sits above 40, a level seen once before, at the top of the dot-com era, against a long-run average near 17. The 10 best-performing big tech names of the past year averaged a 784% gain. The same measure in the year before the March 2000 peak was 622%. The current number is not an echo of the old mania. It is louder.

Then comes the supply. In February, Musk folded xAI into SpaceX. In May, SpaceX filed for what is set up to be the largest IPO in history, a raise reported around $75 to $80 billion at a valuation discussed between $1.75 and $2 trillion. Read the filing the way an investigator reads a bank statement, not the way a fan reads a poster. The xAI side of the business produced roughly $3.2 billion of revenue while burning something near $14 billion of cash. One analyst put it plainly: the financials look reckless. And one reported detail closes the loop on this entire issue: Anthropic agreeing to pay $1.25 billion a month to rent capacity in xAI's data centers, on a contract either side can end with 90 days' notice. The AI giants are renting capacity from each other, on short fuses, and calling it backlog.

Why does the IPO matter to a fraud newsletter? Mechanics. When a giant company lists, the big funds that must buy it need cash, and they raise that cash by selling what they already own, which is mostly the same handful of AI names holding up the index. In 2000, the NASDAQ peaked almost exactly when a record flood of new shares hit the market, and then it gave back roughly 80% over three years. Burry's warning is that this cycle's flood is more concentrated. Not hundreds of small listings spreading the drain, but a few colossal ones, each capable of pulling tens of billions out of the market in a week.

Hold the two warnings side by side, and the file becomes one file. The company at the center of the largest IPO ever is the same company leasing the chips inside the structure Burry calls fugazi. The demand that justifies the lease is the story that justifies the valuation. The valuation feeds the listing. The listing drains the market that prices the demand. Circles inside circles, and at the outer edge of every circle, someone's savings.

In 1999 the rhyme was eyeballs. In 2026 the rhyme is compute. The meter changed. The melody did not.

Five controls for reading a boom built on borrowed certainty.

This newsletter is for informational purposes only and promotes ethical and legal practices. Nothing in this issue is investment, legal, or insurance advice; The Fraudfather is not a financial advisor, and decisions about your portfolio or annuity should be made with a licensed professional. Allegations described here are allegations, not findings of wrongdoing; no regulator has accused any party in the Valor transaction of breaking the law, and the companies involved describe the deal as a standard, fully disclosed financing.