Nearly half of Alphabet's and Amazon's record Q1 profits came from marking up Anthropic equity stakes. When hyperscaler investments cycle back as cloud revenue and get booked as growth, the distinction between revenue and a receivable becomes the only question that matters.
Amazon invested $8 billion in Anthropic. Anthropic committed to spending $100 billion on Amazon Web Services over ten years. In Q1 2026, Amazon booked $16.8 billion in pre-tax gains from marking up its Anthropic equity stake, more than half of its $30.3 billion net income. The gains came from the appreciation of a position in a company that projects a $14 billion loss this year and doesn't expect positive free cash flow until 2029.
Follow the capital. Amazon sends equity to Anthropic. Anthropic sends cash back to Amazon as cloud fees. Amazon books the cloud fees as AWS revenue and the equity appreciation as profit. Two line items on the same income statement, traced back to the same original dollars.
The Other Half
Alphabet recorded $36.9 billion in equity gains the same quarter, driven by its roughly 14% stake in Anthropic. Of Alphabet's record $62.6 billion in net income, approximately $28.7 billion came from marking up that stake. Forty-six percent of the quarter's profit was paper.
Fortune put it plainly: half of Google's and Amazon's "blowout AI profits" came from a stake in Anthropic, not from their actual businesses.
Both companies' AI narratives rest partly on the same asset. Amazon holds the larger stake. Alphabet bought in at a lower cost basis. Both are marking it up based on fundraising rounds that include additional investment from Amazon and Alphabet themselves. The valuation of the asset reflects, in part, the enthusiasm of its own shareholders.
The Backlog
Microsoft's version runs through a different mechanism. Microsoft invested roughly $13 billion in OpenAI for a 27% stake. OpenAI committed to $250 billion in Azure spending over the partnership's lifetime. OpenAI now accounts for 45% of Microsoft's $625 billion commercial backlog.
That backlog number drove commercial bookings to 230% year-over-year growth. The growth justified $190 billion in calendar-year 2026 capital expenditure. The capex builds Azure infrastructure. The infrastructure serves OpenAI. OpenAI's consumption becomes the backlog that justifies the capex.
OpenAI's annualized revenue is roughly $25 billion. Its committed Azure spending dwarfs that figure by an order of magnitude. The difference is funded by investors, including Microsoft.
The Clean Layer
NVIDIA sits above this loop. Its cash economics are different.
Operating cash flow hit $50.3 billion in Q1 FY2027, converting 62% of the quarter's $81.6 billion in revenue into cash. Free cash flow was $48.6 billion. The company returned $20 billion to shareholders in buybacks and dividends in a single quarter and still grew its cash pile.
NVIDIA is fabless. Gross margins run above 73%. Accounts receivable DSO sits between 46 and 52 days. Customers pay on normal terms, in cash.
Customer concentration is the vulnerability. Four customers represent 61% of revenue. Those customers are the hyperscalers running the equity loops. NVIDIA gets paid. The question is whether its customers' customers can keep paying.
The Arithmetic
Anthropic hit $30 billion in annualized revenue in early 2026, passing OpenAI. Inference margins climbed to 70%, up from 38% a year earlier. The company projects its first profitable quarter in Q2 2026: $559 million in operating profit on $10.9 billion in revenue.
But Anthropic itself cautions investors that planned infrastructure spending in late 2026 and 2027 will push results back into negative territory. Internal projections show a $14 billion loss for 2026. Positive free cash flow isn't expected until 2029. The company has committed over $1 trillion to infrastructure over the next several years.
Ed Zitron's May 2026 analysis argues that operating-profit figures shared in a fundraising context exclude costs that would appear in audited filings, including equity-funded compute commitments and stock-based compensation. The profitable quarter is real within a narrow accounting frame. The cash burn is real within a wider one.
A company with $30 billion in annualized revenue, $1 trillion in forward commitments, and no projected free cash flow for three more years is growing. The growth is real. But every dollar of growth still requires more capital than it generates.
Meanwhile, Amazon's trailing twelve-month free cash flow fell 95%, from $26 billion to $1.2 billion. Quarterly capital expenditure hit $44.2 billion. Amazon's $8 billion Anthropic investment is now worth over $70 billion on paper. On paper.
The Terminal Question
None of this is fraud. Corporate venture investing has always created customer relationships. Cisco invested in customers who bought Cisco gear. Intel Capital funded startups running Intel chips. The practice predates the internet.
What's new is scale. Anthropic's $100 billion AWS commitment is larger than most countries' annual technology budgets. OpenAI's $250 billion Azure obligation would have been the single largest commercial contract in history a decade ago. The equity mark-ups flowing through quarterly earnings are large enough to determine whether companies worth over $1 trillion beat or miss estimates.
The terminal question: do AI application companies eventually generate enough genuine end-user revenue to cover their cloud bills without continuous equity infusions?
If yes, the loop resolves. Cloud bills get paid from customer revenue. Equity stakes appreciate on earnings. The circle straightens into a line.
If no, the hyperscalers own equity in companies that can't cover their cloud bills. You can't weather a storm with equity. Bonds give you ownership. But equity in a company that spends more on your cloud than it earns from its users is a receivable that goes to zero.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)