The question of whether artificial intelligence has spawned a speculative bubble has become almost tedious. But Apollo Chief Economist Torsten Slok offers a more nuanced and unsettling perspective: the AI boom is generating real profits, but not in the way the stock market seems to believe. In a blog post published Friday, Slok dissected the AI value chain into four tiers—models and applications, cloud and compute, energy and grid, and silicon and equipment—and found a stark asymmetry. Using data from Pitchbook and Bloomberg for companies ranging from OpenAI and Anthropic to Microsoft, Amazon, Constellation Energy, Nvidia, AMD, and Micron, he calculated that chipmakers and equipment suppliers enjoy a robust 41% operating margin. Meanwhile, companies building AI models and applications, such as Anthropic, are bleeding money with a negative 59% operating margin.

This imbalance, Slok argues, is not a temporary anomaly but a structural flaw. The profits flowing to upstream providers are not being generated by end-user demand for AI services. Instead, they are funded by investors who are betting on future returns. "The AI boom's profits are currently being funded by investors rather than earned from customers," Slok said. "The upstream margins are real, but they are paid for out of capital raised by the layer losing money, not out of cash generated by end demand." This creates a precarious situation where the entire industry's profitability depends on the ability of money-losing AI startups to keep raising capital.

Goldman Sachs now projects that AI investments will swell beyond $1 trillion in 2026. Yet, so far, the technology has little to show for itself in terms of broad economic productivity gains or profit margin expansion outside the Magnificent Seven tech giants. Should the flow of financing dry up, the lopsided margin structure could topple the industry's stability. "The bottom line is that the most profitable part of the AI value chain depends on the least profitable part continuing to grow revenue or raise capital," Slok concluded. "Capital can bridge the gap for a while, but not indefinitely. And therein lies the risk: will the ROI show up for AI's end customers fast enough to sustain the spending that is generating those upstream margins?"

Slok is not alone in sounding the alarm. The Bank for International Settlements, in its annual report published in June, noted that the five major hyperscalers are investing at a pace that outstrips their earnings and free cash flow, forcing them to issue debt. A Bank of America analysis from last November found that in 2025, those same hyperscalers issued $121 billion in debt—four times their average annual issuance over the previous five years. "Disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust, with potential knock-on effects on financial conditions," the BIS warned.

Tech writer Ed Zitron has taken this argument further, using the example of Oracle to illustrate the fragility. Oracle, as of the end of fiscal 2026, has a negative cash flow of $23.7 billion, nearly $130 billion in outstanding debt, and $260 billion in lease commitments for AI infrastructure projects that have not yet begun. Its massive bet is tied to a $300 billion deal with OpenAI signed last September. "Oracle's existence—and Larry Ellison's personal wealth—hinges on whether OpenAI can make good on its promise to spend $300bn in compute," Zitron wrote in a Substack post. He calls this the "most-obvious and under-discussed part of the AI bubble": while hyperscaler capital expenditures in the 13-figure range are fueling a semiconductor boom, there is scant evidence that AI applications will justify the spending. The greater danger, Zitron explained, is not just Oracle failing to deliver, but other tech giants pulling back. "If Microsoft, Google, Amazon and Meta decide that it's time to stop spending $30 billion or more a quarter on GPUs, RAM, storage, and data center construction," he said, "that'll tear a hole in the side of what people assume is a permanent supercycle."