Apollo's Torsten Slok lays out the two-punch sequence that may finally burst the AI bubble — and drag the S&P 500 down with it

“If Chinese models keep gaining and token prices keep falling, the hyperscaler cash flows expected may prove too optimistic,” Torsten Slok, chief economist at Apollo, recently wrote in a market note.

For many years, AI chip customers had generated far more cash than the chipmakers. Now, that dynamic has flipped due to the hyperscalers’ mad scramble for semiconductors to stuff into data centers and other critical AI equipment. This resulted in chipmakers profiting handsomely for the first time, with Bank of America analysts dubbing the phenomenon a “generational transfer in free cash flow.”

Amazon, Alphabet, Meta, Microsoft and Oracle generated a record $260 billion in free cash flow in 2024. However, this pack of hyperscalers is projected to “turn negative“ for the first time this year, Yahoo Finance reports. By comparison, chipmakers Nvidia, Micron, Broadcom and Applied Materials are projected to generate $430 billion in free cash flow throughout the next 12 months, a record for the companies.

Slok raised the possibility of the stock market tipping into correction if hyperscalers don’t reach profitability on schedule or even a full-blown recession, since the “Magnificent Seven” group of tech companies account for an outsized share of the S&P 500. If equity prices take a hit, Wall Street could be in for a rude awakening.

Price war brewing between AI giants

Tokens are the fundamental unit of AI computing — similar to a word fragment — and AI labs have sold them to AI-curious firms. Now, those firms are increasingly leaning towards how they can tighten their belts and rein in token spending.

Last month, the Wall Street Journal reported that OpenAI was weighing major price cuts to increase the appeal of its AI products. The approach is designed to get ahead of Anthropic, which is also expected to slash the cost of their models to better compete. OpenAI executives are already signaling the company’s effort to provide a bargain in the company’s products.

“People are really saying ... ‘My company spent my entire 2026 budget in Q1. Can you make this more efficient?’” OpenAI CEO Sam Altman said in a recent AI summit. “We are continuing to push on that more with models. I think we’ll have a lot of ways we can help people get more value for less spend.”

AI’s kingpins are also staring down another chapter in their competition, one that puts a premium on customizing models for specific tasks and workflows rather than the ability to claim progress on their overall technical performance.

Chinese models gaining ground

Slok observed that Chinese firms account for a larger share of popular AI models, given that they offer many of the same features as those from OpenAI and Anthropic at a fraction of the price.

DeepSeek and Moonshot AI have become popular around Silicon Valley, the WSJ reports. DeepSeek — which broke off from a Chinese hedge fund — shocked the business and tech worlds in early 2025 when it built on the cheap a cutting-edge AI model with fewer cutting-edge chips. Suddenly, America’s lead in the AI race wasn’t as durable as many analysts had forecasted.

What also separates China’s AI landscape is longstanding financial support from the government. Similar to the U.S., Beijing has prioritized AI development for its economic growth and the state has also pushed for its adoption in government and the business sector.