AI's $581B buildout complicates Fed's inflation fight

AI infrastructure spending is pushing up prices for electricity, chips, and software — complicating the Federal Reserve's inflation fight even as Silicon Valley promises deflationary gains ahead.

The Federal Reserve held rates at 3.5% to 3.75% in July, but the $581 billion U.S. AI infrastructure buildout this year is pushing up prices for electricity, chips, and software — complicating the central bank's inflation fight.

"The cost and inflationary aspect is really complicating Kevin Warsh's job," said Peter Boockvar, chief investment officer at One Point BFG Wealth Partners. "He wants to believe in the productivity enhancements down the road — but it's not something he can react to."

Household electricity prices rose 10.1% in the two years through June, outpacing the 6.3% overall increase in consumer prices, according to Bureau of Labor Statistics data. DRAM memory costs are set to rise 400% by year-end versus 2024, JPMorgan Chase estimates, while computer software and accessories prices have climbed 22.9% since June 2024.

The tension between near-term AI-driven inflation and promised long-term productivity gains has split the Federal Open Market Committee. Minneapolis Fed President Neel Kashkari dissented in July in favor of higher rates, citing data-center investment as a new demand element in inflation. Money markets are evenly split between a 25-basis-point hike and a pause in September, according to CME FedWatch data.

Fed Chair Kevin Warsh has staked his credibility on the view that AI will prove disinflationary. In November, before his confirmation, he wrote that "AI will be a significant disinflationary force, increasing productivity and strengthening American competitiveness." He has since appointed Stanford professor Charles Jones — a leading scholar of AI's economic effects, now on leave at Anthropic — and venture capitalist Marc Andreessen to a task force examining how AI will shape the economy.

But the data so far cuts the other way. Goldman Sachs Research projects U.S. AI capital expenditure at $581 billion this year, or 1.8 percent of gross domestic product, rising to 2.8 percent by 2028. Globally, spending could reach $1 trillion in 2026. The buildout has snarled supply chains for semiconductors and data-center capacity, with AI companies buying up every available chip from Nvidia and others faster than manufacturers can ramp production.

Adoption Lags the Hype

Corporate adoption remains uneven, delaying the productivity gains that could offset these costs. A Census Bureau survey published in May found that only 17 to 20 percent of U.S. businesses report using AI, with large firms adopting far more aggressively than small ones. OpenAI chief economist Ronnie Chatterji said the gap between frontier firms and typical companies has widened to eight times in token usage per user, up from two times three months ago.

"The companies that are reorganizing their workflows around it and changing the way they work around AI, they're having more success," Chatterji said.

Julie Averill, former chief information officer at Lululemon, said the technology works but the organizational friction remains. "The things that have always made implementations in large companies difficult still exist, which is people," she said. "Getting people to change their behaviors, taking them along the journey with you, and getting them to trust the model, that's hard."

Economists call this the "weak links" problem. Jones notes that jobs are bundles of tasks, some more amenable to automation than others. Nobel laureate Geoffrey Hinton predicted in 2016 that radiologists would be unnecessary within a decade; instead, their numbers grew as AI made them more valuable by automating only a fraction of their work.

A Split at the Fed

Fed Governor Lisa Cook, speaking in Anchorage on Wednesday, said inflation "is just too high" at 3.7 percent on the personal consumption expenditures index — nearly double the 2 percent target. She cited AI infrastructure spending alongside Middle East energy costs as "two unexpected sources" of price pressure, and said she is "prepared to act by raising rates if necessary."

The last time the Fed faced a similar supply-side inflation shock from a technology buildout was the late-1990s internet boom, when the central bank raised rates through 2000 before the productivity dividend arrived. Boockvar notes that even the internet era delivered only a 1.5 percent average annual productivity gain over 30 years, versus a 2.5 percent average over the past half-century.

"To think that generative AI is going to bring that level of enhancement to the economy, relative to the internet, is tough," Boockvar said.

Warsh has adopted a more cautious tone in recent weeks. In July, he said the "precise timing and magnitude of effects on the supply side remain hard to predict." The task force's findings, expected within months, will shape whether the Fed treats AI as a disinflationary force or a new source of price pressure.

This article is for informational purposes only and does not constitute investment advice.