In his first major policy address as Federal Reserve Chair, Kevin Warsh used the annual Jackson Hole symposium to place artificial intelligence at the center of the central bank's forward-looking framework.

Titled "In Our Time" and delivered on his 100th day in office, the speech framed the current moment as a "historical hinge point." Warsh noted that AI's practical advances have outpaced even the most optimistic predictions made just two or three years ago. Capital is pouring into associated infrastructure as a kind of "super-Moore's law" and shifting scaling laws are redefining both the pace and nature of innovation. The Fed, he said, is "watching closely," acknowledging AI as a new variable with the potential to become a new factor of production, one capable of influencing both the real economy and the execution of monetary policy. Early in his remarks, he explicitly flagged that while the presentation could be read as a roadmap, it should not be interpreted as forward guidance.

Warsh also highlighted the convergence of capital and labor in the creation of large language models, noting a surge in token sales. He cited reports indicating that annualized token sales from two leading laboratories alone have surpassed $100 billion, a more than 500% increase year over year. While the growth trajectory remains steep, he conceded that the timeline and distribution structure are still uncertain and far from settled.

Turning to the core questions guiding Fed research, the prepared text outlined several key areas: whether AI applications will generate significant, sustainable economy-wide productivity gains and when those might emerge; whether token usage complements or substitutes for labor; whether next-generation models will demand higher capital intensity or develop more capital-light pathways; and how market structures will evolve. A crucial unknown is how early surpluses will be allocated among scarce asset owners - such as labs, chip makers, energy providers, and cloud operators - before any benefits trickle down to businesses and consumers, and what that means for workers and the Fed's employment mandate. The equilibrium pricing of tokens also remains unclear: whether frontier models will command sustained premiums and whether older models will fall to near-marginal cost.

A new task force on productivity and employment will lead this analysis. Warsh confirmed he has held preliminary discussions with its leaders and those of four other working groups, but was explicit that their recommendations will arrive later and "will have no impact" on the current policy cycle. He framed the ongoing intellectual work as preparation for future policy conditions rather than near-term action.

Shifting to policy principles, Warsh reaffirmed the 2% PCE inflation target as "firm and fixed." He emphasized that price stability neither happens automatically nor through mean reversion alone, and that delivering it is the Fed's responsibility. He stressed the need for confidence, clear direction, and sufficient speed in returning inflation to target, warning that otherwise "we still have work to do." On financial conditions, he observed little evidence of significant policy restraint in credit and lending markets. Despite stress in sectors like housing and agriculture, he concluded that, on balance, he finds it difficult to characterize broad financial conditions as tight. Markets read these remarks as signaling that rate hikes remain an available tool, running alongside the AI productivity narrative rather than replacing the current inflation focus.

Warsh described forward guidance as a legacy of crisis-era policy that should be constrained in normal times, arguing it risks creating ambiguity under the guise of clarity and constrains the committee's ability to respond flexibly. He cautioned against a "hall of mirrors" dynamic where markets and central banks fixate on each other's expectations, noting that when misjudgments occur, the costs fall first on those without financial assets. He also dismissed mechanical reaction functions, such as Taylor-rule-style formulas, as working better in laboratories than in the field. Unconventional tools, he argued, should be reserved for genuine crises and used cautiously otherwise.

The speech offered no specific guidance on the number of potential rate hikes, did not elevate token sales to official statistics, and announced no new macroprudential tools targeting data centers or laboratories. The timeline for releasing task force recommendations remains unspecified. Whether AI will dampen inflation or reshape the neutral rate of interest remains an open question - not yet a settled new data point on the dot plot.