The accounting trick letting Big Tech pretend to be more profitable than it actually is just got a $160 billion amplifier.
The Summary
- Big Tech companies recorded $160 billion in unrealized gains from paper profits on stakes in AI companies like OpenAI, Anthropic, and SpaceX
- Analysts say these windfalls have muddied tech sector earnings metrics, making it harder to assess actual operational performance
- The gains highlight the volatility and speculative nature of tech-driven financial markets where valuations swing wildly on narrative
- Watch for: How much of Big Tech's "AI leadership" is just mark-to-market accounting on bets they placed on other people's companies
The Signal
Big Tech's earnings reports now include a feature that would make any CFO nervous: massive unrealized gains from stakes in private AI companies that can evaporate as quickly as they appeared. The $160 billion windfall comes from paper valuations on investments in companies like OpenAI, Anthropic, and SpaceX, not from selling products or services. When Microsoft reports earnings, a chunk of the "profit" comes from OpenAI's latest funding round pushing up the value of its stake. When Google posts gains, part of that reflects Anthropic's valuation surge.
This creates a hall of mirrors problem for anyone trying to understand what Big Tech actually earns from AI. Is Microsoft making money from Copilot subscriptions, or from holding OpenAI equity while VCs drive up the price? The distinction matters because one is sustainable revenue, the other is speculative markup.
"Analysts say these large paper windfalls have muddied the tech sector's earnings metrics."
The accounting works like this:
- Big Tech invests in AI startups at Valuation A
- Six months later, new investors come in at Valuation B (higher)
- Big Tech marks up its stake and books the difference as income
- Earnings look great, even if the AI product loses money
The Financial Times notes that analysts are struggling to separate operational performance from these paper gains. When you strip out the unrealized gains, how profitable is Big Tech's actual AI business? Nobody knows, because the numbers are blended. This matters for investors trying to price Big Tech stock, but it matters more for understanding where the AI economy actually is. If the profits are mostly mark-to-market magic on private company stakes, we're in a different phase than if Big Tech were printing cash from deployed AI products.
Crypto Briefing frames this as highlighting volatility and speculation in tech markets. That's true, but it understates the problem. These aren't just volatile assets. They're illiquid private company stakes being valued by venture rounds, not open markets. The $160 billion could be real. Or it could be the consensus hallucination of a funding environment where everyone believes AI is worth infinity, so they price it at half of infinity and call it conservative.
The Implication
If you're evaluating Big Tech's AI strategy, separate the operational AI revenue from the investment markup. Microsoft and Google aren't just AI companies. They're also AI venture funds that happen to report their portfolio gains as operating income. That's not necessarily bad, but it's not the same as building a profitable AI business.
Watch the next down round. If any of these AI darlings reprices lower, Big Tech will have to mark those stakes down, and suddenly earnings will look worse even if nothing changed operationally. The accounting cuts both ways. Right now it's a tailwind. When the music stops, it becomes a drag.