Andy Kessler cautions investors in his column “The Hallucinatory AI Math” that rosy AI revenue forecasts may hide a lurking risk: falling AI token prices. The piece warns that even as AI services scale, cheaper tokens could squeeze margins and reshape earnings expectations.

Kessler frames the threat through the Jevons Paradox, named after economist William Stanley Jevons, who showed that efficiency gains can spur greater consumption. The paradox explains why a drop in coal costs during the industrial revolution fueled deeper demand, a pattern that could repeat in the AI sector.

If token prices collapse, companies could face thinner revenue streams while still investing heavily in data centers and talent. Investors will need to reassess valuation multiples that currently assume steady growth, potentially tightening the market for high‑growth AI names. Such dynamics may compress price‑to‑earnings ratios for firms like OpenAI, Anthropic, and others, forcing investors to seek higher‑quality growth signals.

The commentary underscores that efficiency gains in AI can backfire, echoing historical lessons. Market participants must monitor token pricing trends closely, as they could dictate whether the industry’s projected earnings stick or evaporate. Ultimately, the narrative cautions that AI’s value will hinge on balancing cost efficiency with sustainable revenue streams.