The machines aren't just coming for your job—they're coming for your purchasing power first.
The Summary
- Fed Governor Lisa Cook says AI infrastructure spending, particularly data centers, is straining supply chains and could keep inflation elevated through 2027
- AI could simultaneously improve the Fed's own data analysis capabilities, creating a strange dual reality where the technology both helps and hinders monetary policy
- Productivity gains from AI remain uncertain in timing, meaning we're paying the infrastructure costs now without seeing offsetting economic benefits
The Signal
The Federal Reserve is watching the AI buildout with the same nervous attention it gives oil prices and housing starts. Lisa Cook's warning about data center spending marks a shift in how central bankers think about technology investment. This isn't just another capex cycle. The sheer scale of compute infrastructure required for the agent economy is creating bottlenecks in electrical equipment, cooling systems, and specialized chips that ripple through the broader economy.
The numbers tell the story. Data center construction has moved from a niche industrial category to a macroeconomic force. Every hyperscaler, every AI startup, every company building agents is bidding for the same limited pool of high-voltage transformers, the same fabrication capacity, the same skilled tradespeople who can wire these facilities. That competition pushes prices up across the board.
"AI's dual impact on the Fed highlights a complex balance between leveraging technological advancements for efficiency and managing inflationary pressures."
Here's the paradox: Cook also noted that AI could sharpen the central bank's own analytical work, giving the Fed better real-time data on economic conditions. The institution trying to control inflation caused by AI is simultaneously using AI to understand that inflation. The Fed could theoretically make better policy decisions while those decisions get harder to execute.
The productivity question looms largest. AI-driven investments may strain supply chains while productivity gains remain uncertain in timing. We're in the investment phase, not the returns phase. Companies are spending billions on infrastructure for agents that might deliver massive efficiency gains in 2027, 2028, or 2030. But the concrete, copper, and electricity get bought today at today's constrained prices.
This is supply-side inflation with a technological face. Unlike demand-driven price increases that the Fed can cool with rate hikes, infrastructure bottlenecks don't respond to monetary policy. You can't raise interest rates and conjure more electrical transformers into existence. The buildout has its own timeline, dictated by manufacturing capacity and construction schedules, not the federal funds rate.
Key factors driving the concern:
- Physical infrastructure constraints (power grid upgrades, cooling systems, specialized facilities)
- Competition for scarce technical talent and specialized equipment
- Long lead times on essential components creating persistent supply-demand imbalances
The Implication
If Cook is right, we're entering a period where the AI economy creates its own inflationary pressure independent of traditional economic cycles. For builders in the agent space, this means your infrastructure costs aren't coming down soon. Plan accordingly. For investors, it means the "AI is deflationary" thesis might be premature—the cost compression comes after the buildout, not during it.
Watch how other Fed governors talk about technology investment in the coming months. If Cook's view becomes consensus, it changes how markets price both AI companies and inflation-sensitive assets. The agent economy might be inevitable, but cheap isn't.