US tech stocks fell because record AI profits hide negative cash flow, circular financing and hidden debt.
Usually stocks crash when companies fail, report a loss or slow down growth, but these companies did the complete opposite. Google generated $39 billion in cash from operations and still lost $300 billion in market cap. Amazon lost $120 billion and Nvidia $80 billion without even reporting that day.
On paper Alphabet said it made $112 billion in profit, the highest quarterly profit in Google's history, and operations generated $39 billion in cash, a 41% rise year on year. But it spent $44.9 billion on AI infrastructure, leaving minus $5.9 billion, its first negative free cash flow since 2004.
Dev sells $12,000 espresso machines for $1,000 upfront, the remaining $11,000 payable only once the cafe profits. Hundreds of shops buy and his revenue explodes to $1.2 million. But he has not sold a hundred machines, he has placed a hundred bets that those cafes sell 20,000 cups daily.
Nvidia is in talks to provide a $250 billion financing guarantee so OpenAI can lease a data center built by a SoftBank subsidiary, with total investment expected to exceed $500 billion. It is separately negotiating chip procurement financing up to $350 billion, so exposure could reach $600 billion against $216 billion annual revenue.
Nvidia provides guarantees, SoftBank builds the data center, OpenAI leases computing capacity and uses Nvidia-backed financing to purchase Nvidia chips. Nvidia recognises revenue and orders, then provides more guarantees for even more projects. In that loop Nvidia is simultaneously the supplier, the investor and the debt guarantor.
OpenAI has signed roughly $1.4 trillion in infrastructure commitments while its ARR is $25 billion, and Sam Altman has publicly admitted they lose money even on $200 pro subscribers. If it cuts costs and turns profitable with its jalapeno chips and Sol and Terra models, it becomes the greatest company on Earth.
Stretching the useful life of chips and servers. $30 billion of servers over three years books $10 billion of cost a year; declare six years and it falls to $5 billion, raising reported profit by $5 billion. Meta moved from 3 to 5.5 years, Google from 3 to 6.
It is borrowed through shell companies. Meta's data center costs $29 billion, but a new shell called Binet Investor LLC borrowed $27.3 billion from bond investors due in 2049 and owns it, with Meta merely the tenant paying rent and owning 20%. Five firms carry $1.65 trillion in AI debt.
A credit default swap is insurance on a loan: a bank pays a monthly premium and the seller repays the debt if the borrower defaults. On 27th July 2026, when the Nvidia-OpenAI news broke, the price of insurance on Nvidia's debt recorded its biggest intraday jump, showing investors are scared.
Token costs are going down dramatically. If they keep falling for the next three years as they did in the last three, enterprises will use AI in every operation, OpenAI and Anthropic will become profitable, Nvidia's projections will come true, and it will be the greatest bet the world ever took.
“Usually stocks crash when companies fail, report a loss or slow down growth.”
“Google generated $39 billion in cash from its operations and still lost $300 billion in market cap.”
“So do you realize, Dev doesn't have revenue. He has placed a bet.”
“Now, do you see the money flows in a circle?”
“Nvidia over here is the supplier, investor and a debt guarantor.”
“So for the same phone and the same money, the expense to the company is dependent on what you tell your accountant.”
“America's AI Panic | Has the Market Crash Begun?.” https://www.youtube.com/watch?v=801JPiGAKOE. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/america-s-ai-panic-has-the-market-crash-begun.html.