Federal Reserve Chair Kevin Warsh devoted an entire segment of his speech at the Jackson Hole symposium to artificial intelligence. Rather than providing any interest rate guidance, he focused on AI’s impact on economic growth, corporate investment, and monetary policy decisions.
Capital expenditure growth points to AI infrastructure development
Wash said the long-standing view of "secular stagnation" following the financial crisis is losing its validity. According to him, corporate capital expenditures have grown at an annualized rate of about 9% over the past four quarters—the fastest pace since 2021—with more than half of this growth tied to AI infrastructure development. He is more focused on whether this growth continues to accelerate than on the absolute scale of spending.
Top laboratories generate annual sales exceeding $100 billion.
Warsh said that the pace of AI advancement is faster than even the most optimistic supporters expected a few years ago. He believes that higher potential growth is emerging.
He also provided a set of market data: the annualized token sales of the two leading AI labs have exceeded $100 billion, representing more than a 500% increase from a year ago. In his view, users purchasing tokens to access models has become a significant revenue stream for AI commercialization.
- Corporate capital expenditures have increased by approximately 9% over the past four quarters.
- More than half of this growth is related to AI development.
- Two leading labs generate over $100 billion in annual token sales.
The Federal Reserve views AI as a new variable.
Wash said the Federal Reserve is closely monitoring market developments in the AI ecosystem and has regarded AI as a "new variable," potentially even a new factor of production. This statement implies that AI is no longer just a topic within the technology sector, but may influence the pace of economically sustainable growth and the appropriate level of interest rates.
He raised several unresolved questions, including whether AI will continue to boost overall societal productivity, when such changes might occur, and whether token usage complements or replaces labor. Warsh did not provide answers, but noted that a Fed internal working group on "productivity and employment" is studying these issues.
Add value or continue concentrating in a few areas
Warsh also noted that the gains from AI may not be evenly distributed across the supply chain. A key question is which holders of scarce assets—such as AI labs, chip manufacturers, energy companies, and cloud service providers—will capture the majority of the new value.
The article also cites NVIDIA’s latest earnings report as a real-world reference. NVIDIA reported a record quarterly revenue of $96.2 billion and disclosed future commitments related to AI infrastructure totaling $366 billion. This illustrates that AI investment continues to rapidly concentrate in a few core areas.
Additional information: The article also cites a previous report stating that approximately 95% of generative AI companies are failing, contrasting with the trend of leading chip and cloud infrastructure companies capturing a greater share of profits.