Federal Reserve Chairman Kevin Warsh sees the massive U.S. investment in artificial intelligence as a force that will eventually drive down costs. His colleagues aren’t so sure. The divide surfaced in recent congressional testimony and Fed meeting records. It highlights a deeper debate about whether this technology spending boom acts as a temporary jolt or a lasting economic shift.
Warsh told lawmakers this month that he doesn’t view a one-time rise in prices from AI infrastructure as necessarily inflationary. “I don’t view a one-time change in prices as necessarily being inflationary, because I think there’s a supply response in that way,” he said, according to a Yahoo Finance report. He expects the buildout to boost productivity and wages over time. Short-term price increases? Yes. Persistent inflation? Not if the Fed responds correctly.
But minutes from the Fed’s June meeting paint a different picture. Many officials worried that strong demand for AI infrastructure would keep upward pressure on prices for tech products and electricity. The specific mention of the AI buildout stood out. It marked one of the clearest signals yet that central bankers track this spending as a key variable in their inflation calculus.
And the numbers are staggering. U.S. hyperscalers — the big cloud and tech operators like Microsoft, Amazon, Alphabet, Meta and Oracle — plan to spend between $660 billion and $725 billion on capital expenditures this year. The bulk targets AI data centers, chips, networking gear and power systems. That’s according to a Bloomberg analysis from late April. Projections for 2027 push toward $1 trillion. China, by contrast, spends a fraction. Yet its top models close the capability gap fast.
Warsh first flagged AI as the economy’s most striking feature in earlier remarks. He noted high-tech spending grew nearly 25% in the first quarter while overall equipment investment rose 8%. “At the Fed, we don’t yet know fully the extent to which the economy benefit from AI,” he said in comments shared by Yahoo Finance on Facebook. “Yet it seems inevitable that which we’re now calling AI investment will soon just be called investment.”
This isn’t abstract theory. Data centers consume enormous electricity. Construction crews work around the clock. Chip demand strains supply chains. All of it shows up in price data. Senators pressed Warsh on exactly these points during his July testimony. He pushed back. Supply responses will follow. Productivity will rise. Corporate profits and worker paychecks will improve without locking in higher inflation.
His view draws from historical parallels, though he avoids direct comparisons in public. Past tech waves — think railroads, electricity, computers — delivered deflationary pressure after initial surges. Warsh believes AI follows suit. Increased worker output spreads across the economy. Costs fall. But that takes time. Markets want clarity now.
Investors have taken notice. AI-related stocks sold off recently. Some call it a correction, not a trend reversal. One market observer on X noted the infrastructure buildout accelerates despite the dip: “$700B+ data center spending. Rising compute demand. Growing optical/networking needs.” Until capital expenditures and earnings slow, this looks like positioning risk rather than fundamental trouble.
Others see warning signs. Debt markets may need to finance hundreds of billions more. If stock prices keep falling, companies face higher borrowing costs or forced equity sales. One analyst compared it to 1999-2000 dynamics. “Nobody wants to lend to companies in which their stock is in free fall.” Hyperscalers already burn cash at record rates. Free cash flow erodes as they pour money into unproven returns.
The Fed split adds uncertainty. Officials don’t agree on timing. Will supply gains arrive in 12 months? Two years? Longer? Warsh argues the central bank controls whether temporary price lifts become inflationary. His colleagues highlight sticky demand. Electricity prices. Semiconductor costs. Construction wages. These don’t fade quickly.
Recent coverage reinforces the tension. A Boston Herald article detailed senators grilling Warsh on AI’s inflation effects alongside his contacts with President Trump. Warsh reiterated patience with certain price pressures he deems one-off. A Reuters Breakingviews column from earlier this year gave Warsh credit for spotting AI’s disinflationary potential, even as software valuations adjusted.
Power constraints loom large. Data centers need reliable electricity. Some regions already face shortages. New generation capacity takes years. Cooling systems add another layer. These bottlenecks could prolong inflationary effects. Or they could spur innovation in efficiency. The outcome remains unknown.
Monetization questions hang over the spending. Meta must show returns on its investments or face valuation pressure, one analysis warned. Hyperscalers bet AI will create new revenue streams — better search, personalized ads, enterprise tools, autonomous systems. If adoption lags or results disappoint, the $700 billion starts to look like overreach.
Yet the momentum feels unstoppable. Spending guidance keeps rising. NVIDIA and suppliers report strong order books. Construction permits for data centers multiply. Private credit markets expand to back the projects. One X post described a “GPU debt backstop” and “Project Trinity” linking hardware, power and construction. The AI-backed debt market could hit $7 trillion by 2029.
Warsh’s optimism rests on productivity. If AI makes workers far more effective, output rises without proportional cost increases. Paychecks grow. Profits expand. Prices stabilize or fall. It’s a supply-side story. Many Fed colleagues emphasize the demand side first. Massive capital outlays create immediate economic heat.
The chairman’s Senate testimony offered a nuanced take. Yes, prices may rise over the next year. The Fed will judge if that constitutes inflation. “Whether that’s inflationary or not, that’s up to the Federal Reserve — and we’re going to have something to say about that.”
Markets hang on those words. Rate cut expectations shift with each data point and official comment. Bond traders parse minutes for hints of consensus. Equity investors weigh the long-term promise against near-term cash burn.
Global dimensions complicate matters. The U.S. outspends China by a wide margin on AI infrastructure. Yet Chinese labs narrow performance gaps with fewer resources. Efficiency matters. If American firms waste capital on redundant buildouts, the inflationary risks grow. Smarter allocation could accelerate the productivity payoff Warsh anticipates.
Private sector views diverge too. Bulls point to insatiable demand for compute. Bears question whether current models justify the expense. Enterprise adoption varies. Consumer applications remain nascent. The technology matures. Economic effects lag.
So the debate continues. Warsh bets on eventual disinflation from AI. Colleagues see persistent pressure. The $700 billion — and rising — spend will test both views. History suggests technology eventually delivers gains. The path there rarely runs smooth. Price signals, policy responses and corporate execution will decide the outcome.
One thing seems clear. This buildout reshapes more than data centers. It influences monetary policy, market valuations, energy markets and labor dynamics. The Fed watches closely. So does everyone else.
WebProNews is an iEntry Publication