A circular accounting loop is quietly fabricating the revenue that underpins the entire AI boom, critics say, echoing the dot-com bust.
A circular accounting loop is quietly fabricating the revenue that underpins the entire AI boom, critics say, echoing the dot-com bust.

A multi-trillion-dollar artificial intelligence boom is running on an accounting mechanism that critics call a "round-trip funding loop," where Big Tech giants invest in AI startups and then book the same cash as revenue when it's spent on their cloud services. The practice is inflating profits and hiding a structural flaw in the AI gold rush, with just two unprofitable startups—OpenAI and Anthropic—anchoring over half of the roughly $2 trillion in future cloud commitments held by Microsoft, Amazon, Google, and Oracle.
"AI is clearly an investment bubble," Zoho founder and Chief Scientist Sridhar Vembu said on X. "The justification is that all massive tech waves spark financial bubbles so saying it is a bubble doesn’t negate the tech itself. And this one is the biggest bubble yet."
The scale of the circular revenue is staggering. In the first quarter of 2026, Alphabet posted a record $62.6 billion profit, but $28.7 billion of that came from a paper markup on its Anthropic stake. Amazon mirrored the move, booking a $16.8 billion paper gain from Anthropic that accounted for more than half its $30.3 billion net income. Behind the headline profit, Amazon’s free cash flow cratered 95 percent to $1.2 billion as it spent $44.2 billion on physical data centers.
The mechanism has created a precarious concentration of risk, with Microsoft now carrying 49 percent of its $627 billion future backlog tied to OpenAI alone. Oracle leans harder still, with 54 percent of its $553 billion pipeline riding on the same customer. The pattern echoes the 2001 dot-com bust, when firms like Global Crossing and Qwest Communications swapped fiber-optic capacity to fabricate sales. While that practice was illegal, the current AI loop is fully compliant with accounting rules.
The funding loop follows a simple, legal, and powerful logic. A tech giant like Microsoft makes a headline-grabbing investment in an AI startup, such as its $13 billion commitment to OpenAI. However, the investment arrives largely as cloud credits, not cash. OpenAI then uses those credits to pay for Microsoft's Azure infrastructure, and Microsoft turns around to book that consumption as fresh commercial cloud revenue. The cash never leaves the building.
This loop is sustaining OpenAI, whose annual cloud bill has reportedly surged past $60 billion, more than double its actual revenue of around $25 billion. Anthropic runs a similar playbook with its investors Amazon and Google. In just nine months, the Claude developer spent $2.66 billion on Amazon Web Services, a figure that roughly matched every dollar it earned. The lines between investor, customer, and supplier have become completely blurred.
The second leg of the loop props up Big Tech income statements. Every time an AI startup raises a new funding round at a higher valuation, its corporate backers mark up the value of their investment and drop the unrealized paper gain directly into net income. This practice is responsible for nearly half of Alphabet's and more than half of Amazon's record first-quarter profits in 2026.
The profits are on paper, but the capital expenditures are very real. Amazon's 95 percent free cash flow collapse came as it poured $44.2 billion into building the data centers required to service the AI demand it is simultaneously funding. This disconnect between paper profits and cash flow is a core warning sign for investors questioning the sustainability of the boom.
The bigger problem begins when AI leaves the protected loop and enters a real-world budget meeting. Uber reportedly burned through its entire 2026 AI coding budget by April after engineers racked up monthly API charges ranging from $500 to $2,000 each using Anthropic's tools. The costs became so unsustainable that Microsoft, a primary Anthropic partner, ordered its own staff to stop using the tool. As Nvidia's vice president of applied deep learning, Bryan Catanzaro, admitted, his team now spends more on compute than on human salaries.
The AI market is officially entering its "prove-it" phase, where the central question is no longer about growth, but whether it can ever pay for itself. Mainstream finance is taking notice, with Fidelity's AI bubble framework already flashing two warning signs from Big Tech's Q1 filings: deteriorating earnings quality and unsustainable capital spending. The boom may not get the chance to prove anything if the warning lights keep multiplying.
This article is for informational purposes only and does not constitute investment advice.