The AI Earnings Illusion: Is AI’s Explosive Growth Built On ‘Fake Revenue’?
Adam Taggart | Thoughtful Money® Aug 12, 2026
Summary
Mark-to-market gains on private AI stakes may be inflating reported earnings for hyperscalers like Amazon and Alphabet.
Amazon's roughly $30 billion quarterly earnings included about $16 billion in paper gains from its Anthropic position.
Massive AI capex will spread across future depreciation, potentially making hyperscaler valuations look far more expensive.
Hyperscaler free cash flow is turning negative as debt issuance rises and borrowing costs climb on overinvestment concerns.
S&P 500 first-quarter earnings growth near 20% may mask underlying growth closer to 8% after mark-to-market effects.
Adam Taggart's takeaways
- Top takeaway As host, he frames the central question of whether AI's explosive earnings growth is built on 'fake revenue' from mark-to-market accounting.
Stephanie Pomboy's takeaways
- Top takeaway The guest argues that mark-to-market gains on stakes in AI startups like Anthropic inflate hyperscaler headline earnings, with about 12 of the ~20 percentage points of S&P Q1 earnings growth coming from this accounting item.
- A looming depreciation wave from heavy AI infrastructure spending (data centers) will further pressure long-term hyperscaler earnings.
- She highlights that roughly 12 S&P 500 companies — the cash-rich hyperscalers — are now free cash flow negative, masking weakness across the broader index.
- Rising interest costs from widening hyperscaler debt spreads are another underappreciated headwind on top of the mark-to-market distortion.