Federal Reserve Chair Kevin Warsh sees artificial intelligence as a potential source of stronger productivity and lower inflation over time, but other senior policymakers are increasingly concerned that the technology’s enormous investment requirements are pushing prices higher today. Their concern is not that AI cannot make the economy more efficient; it is that the costs of building the technology may arrive years before its benefits reach ordinary Americans.
The distinction has become increasingly important for the Fed’s interest-rate decisions. In a November 10 speech, Vice Chair for Supervision Michelle Bowman estimated that AI-related price pressures accounted for roughly one-quarter of the excess in core inflation during the year through September, while Governor Christopher Waller warned that the pressures from AI investment were unlikely to disappear quickly.
My comparison of the Fed’s official inflation figures and Bowman’s estimate suggests that AI could be associated with approximately 0.28 percentage point of the gap between September’s core inflation rate and the Fed’s 2% benchmark. That is an illustrative calculation based on a policymaker’s approximate attribution, not an independently measured contribution or evidence that eliminating AI investment would reduce inflation by that amount.
The latest figures reinforce why officials are cautious. The October producer-price report, released Friday, November 13, showed wholesale prices rising 0.3% during the month and 6.5% over the previous year, although those economy-wide figures cannot establish how much inflation AI itself caused.
Why Warsh Sees an Opportunity
Warsh’s argument begins with a fundamental economic possibility: businesses can eventually produce more goods and services with fewer resources when technology makes workers, equipment and processes more productive. If that additional supply exceeds the extra demand generated by higher incomes and investment, inflationary pressure could ease without requiring a substantial slowdown in economic growth.
In his Jackson Hole remarks on August 28, Warsh described AI as a potentially transformative factor of production, raising questions about how rapidly productivity could improve and how widely the resulting benefits would spread. His enthusiasm was accompanied by uncertainty over whether the gains would primarily benefit technology owners, workers or consumers.
By October, Warsh was openly acknowledging the difficulty of the transition. During a discussion at the IMF annual meetings, he recognized that the AI boom was stimulating demand before its longer-term improvements in productive capacity became fully visible, an important distinction in assessing the AI investment boom.
At his October 28 policy news conference, Warsh argued that stronger economic growth did not automatically mean an overheating economy, pointing to productivity improvements that he believed had begun before the current AI wave. He also acknowledged that those improvements were difficult to measure precisely and that the Fed needed more evidence.
I see an important qualification in his position: Warsh is not claiming that the Fed has already proved AI will bring consumer inflation down. He is arguing that technological improvements may change how much the economy can grow without generating additional inflation, while leaving the size and timing of that change unresolved.
The Inflation Cost of Building AI
The immediate challenge is that creating an AI-powered economy requires substantial physical investment. Data centers need electricity, construction workers, cooling systems, high-performance chips, network equipment and financing, often competing for resources used by households and other industries.
Bowman’s November 10 economic policy speech provided one of the clearest official assessments of this pressure. She said tariffs, energy costs and AI-related price increases had each contributed approximately one-quarter of the above-target core personal consumption expenditures inflation recorded through September.
The September PCE release from the Bureau of Economic Analysis put annual core inflation at 3.1%, excluding food and energy. Compared with a 2% numerical benchmark, that leaves a difference of 1.1 percentage points.
Using Bowman’s approximate share, I calculate: 1.1 percentage points × 25% = 0.275 percentage point, or approximately 0.28 percentage point. This is a transparent illustration, not a separate statistical model. The Federal Reserve formally targets 2% inflation as measured by the overall PCE price index rather than its core version, and Bowman’s attribution does not establish a precise causal contribution from AI.
The calculation nevertheless highlights a consequential policy question. If even a portion of persistent inflation reflects rising investment costs rather than an overheated labor market, the Fed must decide whether those pressures will fade naturally or spread across enough industries to justify higher interest rates.
Bowman offered an important counterargument to aggressive tightening. She largely characterized AI-related inflation as relative price shocks concentrated in particular industries and expected some pressures to diminish as production capacity catches up with demand, while warning that financial-market risks and rising borrowing costs could complicate the transition.
Waller Sees a More Persistent Threat
Governor Christopher Waller’s assessment places greater emphasis on the risk that today’s investment pressures will persist. In his November 10 remarks, Waller said Federal Reserve staff had studied inflation associated with AI infrastructure and energy prices, and that the results provided little reason to expect those pressures to fade soon.
Waller described research using purchasing managers’ reports to construct indicators of forthcoming producer-price movements. According to his account, high-technology price pressures appeared in those indicators before becoming visible in official inflation statistics, and October survey information suggested additional increases could follow.
That evidence matters because the Fed cannot rely solely on a forecast of future productivity when current prices are moving higher. A technology investment surge may eventually make production more efficient while simultaneously increasing costs for manufacturers, electricity users and businesses purchasing electronic equipment.
Waller also identified broader inflation pressures beyond AI. More than 70% of core PCE services components had registered annual price increases above 3% in September, he said, suggesting that policymakers were confronting inflation extending well beyond a handful of technology products.
He indicated support for another quarter-percentage-point rate increase in December even if inflation improved, while leaving open a larger move if inflation remained unexpectedly strong. His position illustrates why the inflation policy debate cannot be settled by assuming that technology-driven growth will automatically reduce prices.
Productivity Is Improving, but There Is a Catch
There is measurable evidence that the U.S. economy is becoming more productive. The Bureau of Labor Statistics reported in its preliminary third-quarter productivity release that nonfarm business labor productivity increased at a 3.6% seasonally adjusted annual rate, while unit labor costs declined at a 0.7% annual rate.
Those figures are significant because they show output per hour growing while labor costs per unit of output fell during the quarter. But they do not establish that AI produced those gains, nor do they demonstrate that businesses passed lower costs through to consumers.
A company might use productivity improvements to expand output, increase wages, strengthen profit margins, finance additional investment or lower prices. The inflationary outcome depends partly on which of those choices businesses make and whether competition forces efficiency gains into final selling prices.
Governor Lisa Cook examined that uncertainty in her September 28 AI economy speech. She anticipated modest disinflation from future productivity growth but did not expect those benefits to arrive quickly enough to offset the broader inflation pressures she saw developing during 2026.
Cook also warned against treating every price increase in an AI-related industry as a reason for higher rates. The Fed’s tools influence economy-wide borrowing and spending, making them poorly suited to correcting a temporary shortage of a particular product without affecting unrelated businesses.
This distinction helps explain why the officials’ views are more nuanced than a disagreement over whether AI is economically beneficial. The critical question is whether an investment boom temporarily changes relative prices or generates broader inflation that becomes difficult to reverse.
What This Means for Workers and Households
For ordinary Americans, the debate extends beyond the price of computer equipment. Rapid data-center construction can increase demand for electricity, specialized labor and construction materials, potentially affecting costs in industries that do not directly produce AI services.
Consumers may experience those pressures indirectly through utility bills, electronic products and other goods that depend on the same resources. Yet attributing an individual household’s higher bill to AI would require local evidence about electricity demand, utility regulation, generation capacity and the other factors affecting prices.
The labor-market implications could be even more consequential. In a November 12 speech at Yale, Governor Michael Barr warned of possible serious short-term disruptions as businesses replace some human tasks with AI, particularly where younger or less-experienced workers perform work that can be automated.
Barr also acknowledged evidence that AI is already improving some workers’ productivity, supporting demand in certain sectors and creating new occupations. His concern was that the benefits could become concentrated among powerful technology firms and investors while some workers face displacement or weaker bargaining power.
His analysis raises a question that productivity statistics alone cannot answer: who receives the financial gains from a more efficient economy? Lower operating costs do not guarantee higher household incomes, better employment opportunities or cheaper consumer products when market concentration and unequal access to technology influence the distribution of benefits.
Governor Cook identified a related dilemma for monetary policymakers. If AI causes unemployment to rise because workers’ existing skills no longer match employers’ needs, lowering interest rates may not solve that mismatch and could instead increase inflationary pressure.
For households, the consequences of Fed policy remain more immediate than those technological forecasts. The transmission of higher interest rates into credit card costs and the wider mortgage rate outlook can affect borrowers before productivity improvements reduce the cost of everyday purchases.
That does not mean every Fed increase causes mortgage rates to rise by an equivalent amount. Mortgage pricing also reflects longer-term Treasury yields, market expectations, lender conditions and credit risk, making any specific household outcome dependent on its financing circumstances.
Why the Fed Has Not Settled the Question
The disagreement is occurring against an already restrictive monetary-policy backdrop. On October 28, the Federal Open Market Committee held interest rates at 3.75%–4%, with Governors Lisa Cook and Cleveland Fed President Beth Hammack dissenting in favor of a quarter-point increase.
Those dissents showed that inflation concerns were substantial enough for two officials to favor immediate action, although they did not constitute a formal committee-wide conclusion that AI alone required higher rates. The October policy decision instead left the committee positioned to consider further information before its December meeting.
The complexity is illustrated by Bowman’s own position. Although she identified AI investment as a significant inflationary force, she also warned that tightening policy too aggressively could further damage interest-rate-sensitive industries, particularly housing and smaller businesses.
That is a meaningful counterweight to Waller’s preference for further tightening. Some AI-related shortages may eventually ease through additional investment and production capacity, while higher borrowing costs could restrain both inflationary demand and the investment required to resolve those shortages.
Warsh’s inflation framework also reinforces an important institutional boundary: the Federal Reserve has reaffirmed its 2% inflation objective, regardless of whether technological innovation alters the economy’s productive potential. Optimism about AI is therefore not an announced replacement for the Fed’s inflation target or a commitment to reduce borrowing costs.
The November 16 Test
The next major opportunity to examine the competing arguments comes Monday, November 16, when Warsh is scheduled to participate in the Federal Reserve’s Productivity and Jobs Task Force conference in San Francisco. The official conference program lists discussions involving technology executives, academic researchers, business leaders and labor-market specialists.
Among the scheduled participants are Stripe co-founder Patrick Collison, Stanford economist Nick Bloom and former U.S. commerce secretary Gina Raimondo. Their participation provides an opportunity for the task force to examine technological progress, business adoption and labor-market adjustments from different perspectives, although the conference itself will not determine the next interest-rate decision.
Warsh described his objective at the October press conference as refining his understanding of whether AI productivity gains could arrive rapidly or gradually. The distinction matters because a near-term supply expansion could help contain inflation, while a slower transition would leave current investment costs exerting pressure for longer.
I would judge the evidence through three developments: whether productivity gains spread beyond technology-intensive firms, whether AI-related equipment and energy inflation moderates, and whether workers’ incomes and job opportunities improve alongside efficiency. Sustained progress across those measures would strengthen the argument that AI is expanding economic supply without creating persistent inflation.
Conversely, continued increases in input prices, limited cost savings for consumers and significant worker displacement would challenge the more optimistic scenario. Even strong productivity growth would not settle the question if the resulting gains remained concentrated among a narrow group of businesses and investors.
The central issue is therefore not whether artificial intelligence can ultimately make the United States wealthier. It is whether the economy can withstand the inflationary cost of building that future and whether the benefits arrive broadly enough, and soon enough, to make the transition worthwhile for ordinary Americans.
