Good morning. Fed Chairman Kevin Warsh used his Jackson Hole speech on Friday to focus on inflation. But he also pointed to an unusual metric that could offer clues about whether AI is delivering the productivity gains businesses are betting on: token prices.Tokens are the units used to measure the data AI models process, and many AI companies charge customers based on token consumption. Warsh called AI a potential “new factor of production,” then asked whether customers will pay a premium for tokens from the most advanced models even as prices for older models fall toward marginal cost. That doesn’t mean token prices are becoming a new Fed indicator. Instead, they could reveal how AI economics are evolving. Gregory Daco, chief economist at EY Parthenon, told me that Warsh appears to see token prices as a window into the evolving AI market, offering clues about competition among providers, differences in model quality, pricing strategies, and computing costs. But interpreting those price signals isn’t straightforward. Falling token prices can tell two very different stories. If AI models become more capable while getting cheaper, businesses could generate more output for every dollar they spend, a sign of genuine productivity gains. But if models become increasingly interchangeable, providers could be forced to compete on price. That could signal commoditization and raise questions about whether the enormous capital flowing into AI will generate strong returns. “Pricing power at the frontier—not usage growth—is becoming the real scoreboard for whether AI is creating value or just consuming capital,” Luke Lango, a technology analyst and publisher of Innovation Investor, told CFO Daily. Token use has become a top-of-mind cost concern for CFOs. For companies spending on AI, token prices are an input cost. But lower cost doesn’t necessarily mean higher returns. “For CFOs, the more relevant question is whether AI adoption is generating measurable productivity gains, improving margins, or creating new revenue opportunities,” Daco said. That gets to the larger issue behind Warsh’s comments. The economics of AI will ultimately hinge on how much value it creates and who captures it. Sheryl [email protected] This story was originally featured on Fortune.com ...read more read less