Balancing Growth and Inflation in an AI-Fueled Economy: A 2026 Update

Early this year we published “Five Economic Trends to Watch in 2026,” which examined a potential wave of misery borne from high inflation and unemployment, discussed the impact of artificial intelligence on jobs and the Fed’s policy responses, and considered how local economies will fare while subject to these trends. Now we are taking stock of our projections.

Unexpected Misery from Oil and AI

The first thing to note is that we had not anticipated war in Iran and the misery created by the resultant spike in oil prices. We are impressed by the US consumer’s and the national economy’s resilience thus far in response to the oil shock and its ripple effects. Meanwhile, we are concerned that recent weakness in job growth suggests households and businesses have begun to show strain as oil price volatility persists.

Oil shocks are inflationary, and yet fuel prices are not the only concern for new Federal Reserve Chair Kevin Warsh. Since taking the helm in late May, Warsh has endorsed the view that AI technologies will raise (or are already boosting) productivity. He points to these technology-induced productivity gains as a positive supply shock that will push inflation down closer to the Fed’s 2% target. There is theoretical and historical precedence for this outcome: When economies produce more with less, the cost of doing business and inflation ultimately come down.

Yet, even if we can reasonably expect AI to lead to productivity gains, there are tough-to-answer questions about the timing and magnitude of AI’s benefits and the implications for inflation. Whether AI’s productivity boom is anticipated and priced into investment decisions, Chicago Fed President Austan Goolsbee argues, will be a determining factor in defining the path of inflation and influencing the near-term Fed policy response. Goolsbee notes that most people expect an AI-driven productivity boom, in contrast to the mid-1990s internet-driven growth spurt, which was unexpected. These expectations mean that AI demand and capital expenditures are front-loaded today before much of the productivity gains are realized. Revving the economic engine like this risks overheating and spurring inflation.

While the Fed chair cites AI’s supply-side benefits in taming inflation, other Federal Open Market Committee members have recently noted AI-related demand’s inflationary effects. The latest Fed minutes recorded that “most participants … pointed to scenarios in which, in the context of stable labor market conditions, inflation would remain elevated due to strong AI-related demand, the conflict in the Middle East, or the effects of tariffs.”

There are many links between AI-driven inflation and consumer inflation. Wealth increases ripple through the economy, fueled by a surging stock market as investors anticipate higher future earnings growth. Then there are “unseen” wealth effects stemming from businesses and households expecting higher wages and earnings from a potential productivity boom. Both trends lead to a generalized increase in purchases and prices.

While expanding wealth bumps up prices, the current data suggest that AI’s biggest inflationary risk comes from business investment. These expenditures flow through to consumers via many channels. There are direct impacts, such as higher electronics prices stemming from increased demand for components, illustrated by Apple’s recently announced price hikes. Other effects, such as inflated car prices resulting from elevated chip costs, echo the inflationary issues that we endured during COVID-19.  

There is a muddled relationship between higher construction and land costs and housing inflation. So, when construction and land costs are bid up by data center buildouts, the psychology around housing affordability is often very different from what inflation statistics indicate. People’s perceived misery about the economy may therefore be greater than the official statistics would suggest. We are closely monitoring the data to see if low consumer sentiment turns into consumer action in the form of a reduction in spending or a rise in inflation expectations, both of which the Fed may fear more than actual inflation.

New Fed Chair, New Fed Mandate?

Judging by his statements as Fed chair thus far, Warsh seems to be adjusting the Fed’s goals away from its traditional dual mandate of inflation and employment and toward an inflation-focused mandate. One might argue that a singular focus on inflation will yield the same employment and overall economic benefits as the dual mandate. But in practice this may not be the case.

I have found that a balanced-approach Taylor rule model (light-blue line in the chart below) is, historically, a good benchmark for actual Fed behavior (black line). The balanced-approach model gives higher weight to measures of economic slack and employment conditions, with the view that those factors drive inflation (i.e., a high unemployment rate pulls down inflation). We present two alternative models below: One gives equal weight to slack and inflation (dark-blue line) and another focuses only on inflation (orange line). Following the inflation-only model would have yielded a more accommodative policy in the late 1990s (a period Warsh has been analogizing) but much more contractionary policy in the 2010s, which is consistent with the arguments Warsh made in that period.

All models indicate the Fed should employ tighter policy, although the inflation-only model suggests less tightening is needed. The differences between models can be explained using Goolsbee’s argument cited earlier. The Congressional Budget Office’s current measures of slack suggest that the economy is overheating. Goolsbee would argue that’s because demand is being pulled forward ahead of productivity benefits. No matter what the mechanism, the models show that if you believe the economy is overheating, we need even higher rates than the inflation-only model suggests.

If the 1990s analogy is apt, higher interest rates may be in our future. Former Fed Chair Alan Greenspan, who was an early believer in the internet’s productivity-boosting potential, leaned Fed policy into the internet boom and pushed to keep rates relatively low to spur investment. The Fed eventually raised rates in the late 1990s when the internet boom created what is known as the dot-com bubble and conditions started to get frothy. We use the phrase “relatively low” because, as we noted six months ago, inflation-adjusted rates were in fact quite high during the mid-to-late ’90s. These high rates reflected an economy in which higher productivity and economic growth increased the return on capital. Some current FOMC members have said that AI-driven productivity growth means higher equilibrium interest rates. Meanwhile, Warsh remains focused on AI-driven rate relief.

The forecasts in the policy models illustrated above expect lower inflation in the future, and thus these models all suggest rates should move back to levels close to the current policy setting (the forecasts illustrated in the chart below are based on the FOMCs Survey of Economic Projections). Warsh did not submit a forecast, so his views are not reflected in these projections.

In the coming months, we will stay attuned to the Fed’s navigation of rough seas, as the FOMC strives to implement policy that balances growth and inflation in an environment complicated by zealous AI investment and persistent geopolitical risk.

Increased AI Dependency

Zooming out from our focus on inflation and monetary policy, the entire nation’s near-term economic outlook is largely dependent on AI. As the chart below illustrates, investment in information technology hardware and software is responsible for most of the country’s recent growth. While data center construction is one of the fastest-growing components of the economy, the dollar values are still quite small: roughly $50 billion of a $32 trillion economy. This portion translates to an additional 0.05 percentage points to gross domestic product growth, which would be nearly imperceptible on the graph. For those surprised by the modest construction investment numbers, note that the lion’s share of the eye-popping capital expenditure figures for data centers go toward hardware — the chips, servers and fiber — which is at an annualized $800 billion run rate, up from less than $500 billion two years ago.

IT-energized growth has a precedent in recent US economic history. During the 1990s internet boom, IT made contributions to GDP growth that look like today’s AI-backed economy. Yet economic expansion was less dependent on IT in the 1990s than it is today, since the overall growth rate was higher then, driven by a broadening labor force and growing productivity. The questions we posed at the year’s outset are even more relevant now: Are we in the early stage of a multiyear AI-supported period of growth? Or are we in the bubble stage, analogous to the 1999-2000 dot-com bubble, with a coming slowdown in investment? Are tech companies overvalued today?

Updated EMA Forecasts

As part of our American Growth Project, we analyze 150 microeconomies representing nearly 90% of US economic activity, producing forecasts, unique economic indicators, and translational research. As we update our economic forecasts for the 150 largest Extended Metropolitan Areas, the economy’s dependency on AI poses an analytical challenge.

Many of the country’s fastest-growing EMAs have large and growing tech sectors, which are heavily dependent on AI investment. Will AI expenditure continue to accelerate growth in these EMAs? Or does the reliance on tech raise the risk of contraction if there is a slowdown?

As the graph below illustrates, large EMAs tend to have the highest tech exposure. Yet in many of these EMAs, there is limited capacity for expanding the sector’s physical footprint. In a recent study we identify midsize EMAs where AI-driven growth can scale and discuss the constraints that shape this growth. Madison, Wisconsin, and Albuquerque, New Mexico, are perhaps unsurprising AI growth leaders, given their existing high-tech exposure, while Charleston, South Carolina, may seem an unlikely entrant on the list for its potential for AI-lead growth. Meanwhile, Des Moines, Iowa, and Cape Coral–Fort Myers, Florida, combine strong construction capacity with budding tech industries. These markets are particularly well positioned to support the physical expansion of AI infrastructure.

Returning to our forecasts, data for the first half of the year point to a somewhat stronger economic outlook than anticipated in January. Average projected GDP growth across the 150 EMAs increased from 1.1% to 2.1%. The 100 midsize EMAs saw a larger upward revision, with average GDP forecasts increasing by 1.1 percentage points, compared with 0.6 percentage points for the top 50. In total, all our 150 EMAs are now expected to grow in 2026, and we have upgraded our forecast for 140 of them.

The biggest gainers are generally EMAs concentrated in Southeast. Tallahassee and Jacksonville, Florida, got the biggest boost in the midsize and large city categories, respectively. For the 10 EMAs that experienced growth downgrades, the magnitude was relatively small. Underlying economic dynamics, including stronger-than-anticipated productivity growth, were partly responsible for the upgrades. Yet our midyear forecast update also incorporated significant methodological improvements, making it difficult to attribute the changes to one specific contributing factor.

Despite the methodological changes, our fastest-growing large EMAs remain largely consistent, perhaps an indication of the robustness of our modeling. Austin remains our expected growth champion, followed by Nashville (previously No. 6), Seattle (5), Raleigh and Durham (4) and San Antonio (8). The San Francisco Bay Area and Salt Lake City both dropped six spots, while Tampa moved up by the same amount to join the top 10.

Most of these EMAs have significant tech exposure, which creates an opportunity but also a risk. Labor force growth continues to slow nationwide, limiting the pace at which many regions can expand their workforce. With a steady or declining labor pool, local economies must enhance productivity to drive growth. These trends raise an important question: How can microeconomies boost their productivity as labor becomes more constrained?