Chamath Palihapitiya says the AI boom may be obscuring a costly mistake. On a recent episode of the All-In podcast, he argued that only 0% to 2% of the S&P 493’s recent earnings growth comes from AI productivity, with the rest tied instead to inflation-driven pricing power and aggressive share buybacks.
That claim lands now because the spending frenzy is only getting bigger. Hyperscalers are committing hundreds of billions of dollars to data centers, Nvidia cannot make AI chips fast enough to satisfy demand, and the businesses buying all that capacity are still struggling to show that it pays off. Since generative AI entered the mainstream, the S&P 493 has produced roughly 9% earnings-per-share growth, which Palihapitiya said should not be mistaken for an AI dividend.
He framed the issue as a basic test of capital discipline. AI adopters, he said, should have to prove the technology can generate returns above the risk-free rate, or explain why the money was not simply left on the balance sheet. In his view, the burden is on the buyers, not the vendors, because the infrastructure race has already created a market where spending can look like progress even when the books do not show it.
The friction point is that the scoreboard remains weak. The PwC 2026 CEO Survey found that 56% of CEOs saw no AI revenue or cost benefit, while only 12% said they gained both higher revenue and lower costs from AI. That sits uneasily beside the hundreds of billions flowing into chips and data centers, and it suggests the AI boom is still being rewarded more for promise than proof.
The S&P 493 in this debate is the S&P 500 excluding the largest technology companies driving the AI boom, which makes the earnings question sharper rather than softer. If Palihapitiya is right, the market is treating inflation and buybacks as if they were evidence of AI productivity. The next test is whether Corporate America can show that AI investments are producing enough real return to justify the capital already committed.
