Fed Chair Kevin Warsh gave his first Jackson Hole keynote, spending a lot of time on AI token economics. His speech happened during the yearly event, titled “Financial Innovations: Implications for Payment and Policy,” which saw central bankers and experts meet to chat about digital finance and monetary policy.
Jackson Hole turns to token economics
Warsh noted that annualized token sales from the two leading AI labs have now exceeded $100 billion, a figure that has grown more than 500 percent year-over-year.
The Fed chair framed this as “a sign of the extraordinary capital being deployed in the AI sector,” while cautioning that the economics of this new market remain deeply uncertain. “It’s not obvious where the returns on capital will land or on what timescale,” Warsh told the assembled central bankers.
Frontier versus commodity pricing
Warsh also revealed that the Fed is studying a quite critical question: whether the AI token market will develop a pricing split, similar to how other technology markets have evolved.
Now, the central bank is examining whether frontier-model tokens will command premium pricing due to their superior capabilities, while older-model token prices eventually fall toward their marginal cost of production.
This observation tracks with how token pricing has already begun to stratify. Also, industry analysts have noted that token prices span a wide range: from roughly $1 per million tokens for commodity inference to $150 per million for real-time interactive services. The price span reflects differences in speed, quality, and model capability.
Moreover, the question Warsh raised is whether this stratification will persist or whether competition will compress margins across the board.
Furthermore, some analysts have warned that rising token costs could pressure the AI trade, with inference subsidies ending and companies warning that AI budgets are being exhausted faster than expected.
Macro implications of the AI buildout
Warsh also noted that more than half of this year’s capital expenditure growth can likely be attributed to AI development.
The scale of investment in data centers, Graphics Processing Units, and related infrastructure is now large enough to show up in macroeconomic data.
A few economists have mentioned that while AI spending might boost productivity and cut inflation down the road, in the short run, it’s driving up chip prices and adding to today’s inflationary pressures.
