Kevin Warsh has a point on AI and inflation

LONDON, March 4 (Reuters Breakingviews) - Kevin Warsh claims artificial intelligence will justify lower interest rates. That is a mightily convenient view for the person nominated by President Donald Trump to be the next chair of the Federal Reserve. The former Morgan Stanley banker is still onto something, though. Letting economic orthodoxy constrain genuine growth would be a mistake.

The debate among economists over the likely impact of AI has been raging for years. ​In the United States, the technology may already be helping workers be more efficient. Revised data from the Bureau of Labor Statistics imply hiring slowed far more than initially thought in 2025 even as GDP ‌stayed strong, meaning workers produced more per hour. Sectors that have embraced AI, notably telecom operators, broadcasters and publishers, have raised productivity more than hotels and restaurants, where the technology has fewer obvious applications. The Kansas City Fed cautions, opens new tab that direct AI use remains patchy, but industries that expect a big role for large-language models seem more willing to limit headcount.

The Reuters Iran Briefing newsletter keeps you informed with the latest developments and analysis of the Iran war. Sign up here.

Rapid adoption of AI would be different from the personal computer revolution, which took until the 1990s to deliver efficiency benefits. Back then, Warsh recently recalled in a YouTube interview, opens new tab with Aven CEO Sadi Khan, Fed Chair Alan Greenspan “sat on his hands” in the ​expectation that a growth boost would not push up prices. He was right. Output per hour grew at an average rate of 2.7% between 1994 and 2004, even as inflation fell to 1.9%, echoing a similar innovative period in ​the 1960s.

Warsh thinks the U.S. is due for another technological windfall. AI will "be a significant disinflationary force,” he wrote in November, opens new tab in The Wall Street Journal. Investors are warming to that view. ⁠The recent software selloff reflects anticipation that their clients will use tools like Anthropic’s Claude Code to develop those functionalities themselves.

The problem with Warsh’s belief is that it runs counter to mainstream economic theory. This holds that faster trend growth raises the “natural rate of ​interest” as firms and households bring forward spending. If the central bank does not raise borrowing costs in lockstep, demand can overheat. In this scenario any disinflation caused by technological progress is temporary.

Indeed, Fed estimates depict the natural rate rising when productivity accelerates. Some ​of the central bank’s top officials, such as Lisa Cook, opens new tab, Michael Barr, opens new tab and Philip Jefferson, opens new tab, seem to be leaning the same way. Their views could limit Warsh’s wiggle room, especially as his politically convenient thesis may make it harder to prove his commitment to defending the central bank’s independence. Even Greenspan’s dedication to tolerating technology-led growth had limits, opens new tab, since the Fed kept policy quite tight as the telecom‑led investment boom got going in the mid-1990s and only cut interest rates after hedge fund Long Term Capital Management imploded in 1998.

Still, obsessing about an unobservable natural rate is a mistake. Surveys show companies rarely cite ​borrowing costs as decisive for investment, and households base spending mainly on their income and age. Meanwhile, figures from the Bureau of Economic Analysis stretching back to 1977 show a clear historical correlation between sectors that experience faster productivity gains and those that ​raise prices less. In the specific case of AI, fresh cross-sector research, opens new tab from the euro zone finds that, when the share of firms using it rises by 10 percentage points, producer inflation falls by up to 0.6 percentage points.

The catch is that, while clothes and smartphones have become ‌cheaper, healthcare, education ⁠and housing have become more expensive. This is a pattern known as “Baumol disease.” The American economist William Baumol showed in the 1960s that as productivity rises in tradable industries — those producing goods and services that can be consumed far from the point of production — income flows to activities that are hard to automate or transport, so their prices and share of employment increase. In a paper, opens new tab he co-authored last year, Nobel laureate Michael Spence estimates that nontradable sectors now make up 77% of U.S. jobs and 68% of value added.

Using his approach to classify output and employment reveals the limits of the 1990s boom. Productivity in tradable industries surged, averaging 4.4% growth, but the impact on non-tradable sectors was small. The number of jobs in tradable industries grew only 6% from 1994 to 2004, ​as tech firms shrank after the dotcom bubble burst, while ​employment in nontradable areas expanded 19%. In the former, workers’ ⁠share of value added fell from 62% to 57%, showing how investment-led growth skews gains toward owners of capital. It appears to support the view of the information technology wave as a one-off bump, opens new tab that mostly just benefitted the makers of hardware and software. Adopters of new digital systems like giant retailer Walmart gained in the U.S., but not in Europe.

A better takeaway, however, is that ​the macroeconomic environment matters. In the 1990s, the American labor market was far tighter than the European one, having quickly recovered from a recession, so U.S. firms had incentives to ​squeeze more output from existing workers. That ⁠tradable productivity growth plummeted to 1% in the 2010s suggests that it was the global financial crisis that killed the boom, not digital saturation.

Fast-forward to today, and hiring has also slowed after a growth period. Spending could weaken further if AI widens income inequality, since the wealthy consume a smaller proportion of their earnings. In that case, monetary policy should be looser. If stock market doomsters are right and industries like law, education and health care are able to automate key processes, the deflationary effect will be even more pronounced.

Of course there ⁠are big differences. ​The 1990s were a time when globalization weighed on prices. Now, tariffs are pushing up prices of tradable goods in the U.S. and the data center ​investment boom is straining power, land and supply chains. All these factors place a floor on official borrowing costs.

Still, it would be a mistake for the Fed to try to restrain the AI boom to fit a textbook model, or to assert its independence from an interventionist president. In that sense, Warsh has ​a point.

Follow Jon Sindreu on X, opens new tab and LinkedIn, opens new tab.

For more insights like these, click here, opens new tab to try Breakingviews for free.

Editing by Peter Thal Larsen; Production by Pranav Kiran

• Suggested Topics:
• Breakingviews

Breakingviews
Reuters Breakingviews is the world’s leading source of agenda-setting financial insight. As the Reuters brand for financial commentary, we dissect the big business and economic stories as they break around the world every day. A global team of about 30 correspondents in New York, London, Hong Kong and other major cities provides expert analysis in real time.

Sign up for a free trial of our full service at https://www.breakingviews.com/trial and follow us on X @Breakingviews and at www.breakingviews.com. All opinions expressed are those of the authors.

Share

• X

• Facebook

• Linkedin

• Email

• Link

Purchase Licensing Rights

Jon Sindreu

Thomson Reuters

Jon Sindreu is the London-based global economics editor for Breakingviews. He was previously a reporter and a columnist for the Wall Street Journal, where he covered macroeconomics, financial markets and aviation for 11 years. He holds a master’s degree in financial journalism from City St George’s, University of London. He also holds degrees in computer science and journalism from Universitat Autònoma de Barcelona, in his natal Catalonia.

• Email