Kevin Warsh, who was sworn in as the Chairman of the Fed Reserve in May this year, describes AI as a “structurally disinflationary” force in the economy. He argues that AI will lead to an enormous productivity boom that will lower business costs and expand output, allowing the economy to grow faster without adding to price pressures. When that happens, it should create room for the Fed to cut interest rates.
Right now, that argument is unproven. If anything, the AI infrastructure boom is feeding the inflation that central banks around the world are struggling to contain. While its inflationary impact might pale in comparison to robust consumer spending and the spike in oil prices since the start of the Iran war, the AI buildout is driving up costs for IT talent, electricity, construction, and a wide range of computer components.
Indeed, Fed Governor Lisa Cook recently said that the AI buildout is adding inflationary pressure in the short term and names it as one of the factors delaying inflation’s return to the Fed’s 2% target. While she agrees with Warsh that AI could unleash an era of abundance and hyper-productivity, Cook suggests that the timing and magnitude of the disinflationary effects are uncertain.
AI is inflationary for now
With US tech investment hitting 4.9% of GDP in the first quarter of 2026 – above the Dot-Com peak, according to BCA Research – we are still in the early stages of the AI revolution. Any disinflationary benefits may only surface once the infrastructure is built and the productivity gains permeate through the wider economy.
Even then, many economists have reservations about whether AI will bring about a material structural shift that lowers interest rates and inflation. In a Financial Times and Chicago Booth poll, almost 60% of 45 economists expected AI to shift inflation and the neutral rate by less than 0.2 percentage points over two years. About a third thought the neutral rate could rise.
Warsh’s bullish outlook for AI-fueled disinflation is partly inspired by Alan Greenspan’s late-1990s call to hold off on hikes while productivity surged. Greenspan’s “productivity gambit” was a gamble that technological productivity would prevent a booming economy from sparking inflation. As such, it is helpful to look back on the 1990s internet-driven productivity boom for insight into how a disruptive technology may affect inflation.
Greenspan was celebrated at the time for engineering a soft landing. However, many economists now argue that it is difficult to disentangle how much the tech boom contributed to the disinflation of the 1990s compared to falling oil, metals, imports and farm prices. There is some consensus that the tech boom was an important supply-side force, but it operated alongside several other powerful dynamics.
The 1990s offer mixed lessons
As the internet shows, some goods and services becoming much cheaper does not mean deflation or disinflation across every product tracked in the CPI basket. Software, recorded media, travel bookings and financial services might have become cheaper due to the transparency and efficiencies of digital commerce. But inflation rates for housing, healthcare, education and restaurants were largely untouched.
Furthermore, while prices of some goods and services tumbled from the late 1990s through to the 2010s, they eventually stopped falling. Once a price stabilises, it no longer drags the inflation index down. In some cases, prices can even start to rise after consumers have become accustomed to them falling. We are seeing prices for nearly anything with a computer chip rise right now due to the demand from AI infrastructure companies, and in particular, the prices for end-user products such as smartphones and other computing devices have experienced significant price increases.
There could be nuances in the AI boom. Warsh may be right that the current inflationary effects of AI are transitional and will fade once the infrastructure is in place. It may be that automating white-collar and operational tasks with AI will permanently reduce production costs across many parts of the economy. Companies could pass these savings to consumers, dragging down services-sector inflation, a major component of CPI and PPI.
History repeating or rhyming?
But on the other hand, AI infrastructure is physically hungry, requiring large numbers of chips, power, and construction labour. There are concerns that data centre assets wear out faster than most other capital assets. The AI infrastructure buildout will not be a once-off expense, and inflation has been above target for five years, which leaves less room for error than in the 1990s.
It is also unlikely that AI can dispel the importance of macroeconomics. Labour-market slack, geopolitics, and supply-demand dynamics across oil, metals and other commodities will continue to shape inflation in a post-AI world, just as they did in the 1990s, and how central banks and governments respond to inflation will matter at least as much as technology and macro forces.
A sobering thought on Greenspan’s legacy: He was hailed as a “maestro” for the soft landing. But many of his critics see the seeds of the Dot-Com Crash and the Global Financial Crisis in the loose policies that preceded them. By keeping monetary policy accommodating, the Fed may have fueled the Dot-Com bubble, which eventually burst in 2000. One hopes the same mistake will not be repeated during the AI boom.
