Chamath Palihapitiya Says the AI Boom May Be Hiding Biggest Capital Allocation Mistake in History

Artificial intelligence has completely changed the narrative for this decade. The four largest hyperscalers alone are guiding to roughly $725 billion in combined 2026 capital expenditures, up about 77% from $410 billion in 2025, with the overwhelming majority directed at data centers and AI chips. NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) cannot manufacture AI chips fast enough to satisfy demand, and companies across nearly every industry are racing to deploy generative AI.

Investors have rewarded those building the infrastructure. Yet infrastructure spending is only half the story. The harder question is whether the companies buying AI are generating enough financial returns to justify the investment. According to venture capitalist Chamath Palihapitiya, that answer may soon determine the next stage of the AI boom.

The Productivity Numbers Aren’t Matching the Spending

On a recent episode of the All-In podcast, Palihapitiya made the case that AI return-on-investment is now the defining question for corporate earnings. His argument was not that AI has failed. Rather, he challenged investors to draw a sharp line between the companies selling AI and those buying it.

Strip out NVIDIA, the cloud providers, semiconductor equipment manufacturers, and other AI infrastructure leaders, and what the rest of corporate America has actually earned from its AI investments tells a very different story. On a separate episode earlier in the year, Palihapitiya set an even more specific deadline: he argued companies have roughly 500 days to prove AI spending is translating into measurable economic returns, or Wall Street will stop giving them credit for future benefits that have yet to appear.

The S&P 493, meaning the S&P 500 excluding the largest technology companies driving the AI boom, has produced roughly 9% earnings-per-share growth since generative AI entered the mainstream. Yet Palihapitiya estimates that only about 0% to 2% of that growth stems from AI-driven productivity. The remainder reflects inflation-driven pricing power and aggressive share buybacks rather than genuine operating improvements.

That distinction matters because AI spending continues accelerating while measurable productivity gains remain elusive. Gartner forecasts total worldwide AI spending will reach $2.7 trillion in 2026, a 49.5% increase year over year. The money is flowing in; the returns are not following at anything close to the same pace.

The Data Suggests CFOs Are Losing Patience

Consider how the investment boom compares with the financial results.

The PwC 29th Global CEO Survey, which polled 4,454 chief executives across 95 countries, reinforces Palihapitiya’s concern directly. More than half of CEOs reported AI had neither increased revenue nor reduced costs. Only 12% experienced both outcomes at once. PwC Global Chairman Mohamed Kande described 2026 as “a decisive year for AI,” noting that a small group of companies are already converting AI into measurable financial returns while many others remain stuck in the pilot phase.

The industry even has a name for this phenomenon: pilot purgatory. Companies successfully demonstrate AI in contained pilot projects but struggle to deploy it broadly enough to produce measurable financial gains. Meanwhile, spending has migrated from experimental innovation budgets into core operating budgets, placing AI investments under the scrutiny of chief financial officers rather than innovation teams.

The Burden of Proof Is Changing

Every transformative technology has gone through a period where spending arrives before profits. The internet, cloud computing, and smartphones all required years before productivity gains surfaced across the broader economy. That historical precedent gives AI bulls a reasonable foundation for patience.

Palihapitiya’s argument is not that AI belongs in the failure column. His point is that capital has a cost. If AI spending continues at the pace the hyperscalers are signaling, those investments eventually need to generate returns above the risk-free rate available from Treasury securities. Companies that cannot clear that bar would have been better off leaving the cash on their balance sheets or returning it to shareholders.

That is an uncomfortable conversation, because investors have spent the past three years focused on AI’s astonishing capabilities rather than its financial output. Capabilities alone do not determine shareholder returns. Earnings growth, free cash flow, and return on invested capital do.

Key Takeaway

The AI investment story is entering a more demanding phase. Building powerful models and deploying chatbots impressed investors during the first wave. The second wave will demand proof that AI expands margins, lifts productivity, and generates measurable earnings growth.

That pressure does not spell trouble for AI leaders like NVIDIA or the hyperscalers, whose revenues continue reflecting strong infrastructure demand. For the thousands of companies spending billions to adopt AI, however, the spotlight is shifting fast. The question is no longer whether AI works. The question is whether it earns more than it costs.

Ultimately, the companies that can answer that question with hard financial results are likely to produce the next generation of market winners.

Editor’s note: This update added the four largest hyperscalers’ combined 2026 capital expenditure guidance of roughly $725 billion (up 77% from 2025), Gartner’s September 2026 forecast of $2.7 trillion in total worldwide AI spending, Palihapitiya’s “500 days” ROI deadline from a May 2026 All-In episode, the Menlo Ventures source for the $37 billion enterprise GenAI spending figure, and PwC’s survey sample size of 4,454 CEOs across 95 countries.

Contact [email protected] for any questions or corrections.