The Two AI Economies Inside the Fed

Federal Reserve Chair Kevin Warsh has made his view of artificial intelligence plain ^[1].

a significant disinflationary force — Kevin Warsh

The minutes from the Federal Open Market Committee meeting he chaired in June — released days later — tell a different story. They identify AI infrastructure spending as likely to keep pushing up the cost of electronics and power ^[2].

Many participants noted that ongoing strong demand for AI infrastructure would likely sustain upward pressure on prices for technology products and electricity. — Federal Reserve System

The same man presides over both documents. The contradiction is not a slip. It is a map of the race the Federal Reserve now finds itself watching — and has positioned itself to react to — without being able to control which side wins. The inflationary side of that race is already visible in three channels, and industry leaders say none of them will ease before 2028 at the earliest. Memory chips are the most immediate: hyperscaler demand for high-bandwidth memory has drained the supply of DRAM, forcing Apple to raise laptop and iPad prices 15 to 25 percent and Microsoft to add $100 to the Xbox ^[3]. Samsung has warned the shortage will worsen through 2027, with one executive stating that supply falls far short of what customers need ^[4]. Intel’s CEO estimates relief in 2028; SK Group’s chairman says 2030. Electricity is the second channel: data center demand helped drive PJM Interconnection capacity prices up more than 1,000 percent, pushing industrial rates up 31 percent in Pennsylvania and 26 percent in Ohio ^[5]. The third is skilled labor: data center construction is pulling electricians away from homebuilding at double the standard wage, with roughly 20,000 electricians retiring annually and no adequate backfill ^[6]. These are not policy shocks that can be reversed with a tariff rollback or a diplomatic deal. They are competition for finite physical inputs — chips, watts, workers — and the competition is accelerating. Amazon has projected $200 billion in capital expenditures for 2026, Alphabet raised its guidance to as much as $190 billion, and Alphabet’s CFO has said spending will rise significantly further in 2027 ^[7]. But the deflationary side of the race is just as real, and it is running simultaneously. AI-related layoffs reached 87,714 in the first five months of 2026, surpassing the combined total for 2024 and 2025 ^[8]. Andy Challenger, whose firm tracks job-cut announcements, put it plainly.

AI is now the leading reason companies give for cutting jobs. — Andy Challenger

The June inflation data, released Monday, captures both forces at once. Headline annual inflation cooled to 3.5 percent from 4.2 percent in May, driven down by used cars, clothing, and gasoline. But core inflation rose to 2.6 percent, above the Fed’s target, and the AI-driven components — electronics, electricity — did not ease ^[9]. The number that improved was not the number AI is pushing on. This is the split that makes the Fed’s position so precarious, and it is the split Warsh has prepared for in ways that go beyond any single meeting’s minutes. He has removed forward guidance — the practice of signaling the likely path of rates — which means markets and businesses can no longer count on the Fed to telegraph its next move. He has also shifted the central bank’s center of gravity away from the employment side of its dual mandate and toward price stability, stating plainly that he rejects the old trade-off between inflation and jobs ^[10].

I don't share the view that was expressed a few generations ago that you're going to have to decide whether you're willing to tolerate higher inflation to put more people at work. — Kevin Warsh

Each of these changes makes the same bet in a different way: that the Fed will need to move, and move fast, and that the luxury of signaling patience is one it can no longer afford. Fed Governor Christopher Waller has already said another hot core inflation reading could trigger near-term tightening ^[3]. The central bank is formally weighing rate hikes ^[11]. Warsh has not resolved the contradiction between his public optimism about AI and his institution’s formal warning about it. He has instead built a Federal Reserve that does not need to resolve it — only to be ready when the data picks a side. The instruments that once let a Fed chair signal patience are gone. What remains is a central bank positioned to move before anyone knows which AI economy is the real one.