AI Investment First, Disinflation Later: How Data Centers Are Rewriting the Inflation Path

The largest technology companies are spending several hundred billion dollars a year on data centers to train and run artificial intelligence. Supporters argue that AI will eventually make workers and businesses far more productive, and higher productivity tends to slow price increases. That may prove true. But the order of events matters, and it's lopsided: spending happens now, while productivity gains remain a forecast.

In the short run, data center construction behaves like any other demand surge. It competes for electricity, equipment, land, and skilled labor, and it pushes up the prices of all four. Any broad slowdown in inflation from AI depends on the technology spreading through the rest of the economy, a process that historically takes years.

That sequence, investment first and disinflation later, creates a real problem for the Federal Reserve. The central bank must set interest rates based on pressures it can measure today, while the relief that might justify lower rates sits somewhere in the future. How the Fed handles that gap will shape borrowing costs for the rest of the decade.

Inflation is the rate at which prices across the economy rise over time. The Federal Reserve targets inflation of 2 percent per year, a pace slow enough that households and businesses to plan around. The Fed's track record is mixed: inflation reached about 9 percent in mid-2022, the highest in four decades, before falling back.

Inflation is a rate of change, so it is measured against something. That something is the Price level, the average cost of a broad basket of goods and services at a point in time. Statistical agencies estimate it through indexes such as the Consumer Price Index and the Personal Consumption Expenditures index, the Fed's preferred gauge. Inflation is simply how fast that index climbs.

The distinction matters for this argument. A jump in one price, such as electricity, raises the price level a little but does not by itself produce sustained inflation. Inflation becomes a problem when price increases spread across many categories and persist.

Two more terms are needed. Disinflation means inflation is slowing: prices still rise, but more gently. Deflation means the price level actually falls. The optimistic case for AI is a disinflation story, not a deflation story. Even in that scenario, prices keep rising, just more slowly.

AI is stepping out of the digital screen and entering the physical world. Robotics VC deal activity hit all-time highs in early 2026, with investments climbing toward sixteen billion dollars across more than four hundred and fifty deals. 

Driven by manufacturing labor shortages, aging workforces, and explosive tech breakthroughs from NVIDIA, Tesla, and a host of specialized component suppliers, automation is accelerating across factories and warehouses globally. 

Whether you are looking at AI compute, machine vision sensors, motion control actuators, or humanoid robotics, the infrastructure buildout is happening right now. Download our comprehensive guide to discover the top investable robotics sectors and market drivers.

Download Robotics Guide Free

Economists distinguish between the demand side and the supply side of an investment. On the demand side, building a data center means buying concrete, steel, cooling systems, electrical transformers, chips, and paying electricians and construction crews. That spending happens immediately. On the supply side, the finished facility adds productive capacity, but only once it is running and only to the extent its output makes other businesses more efficient.

Electricity is the clearest channel. A 2024 Lawrence Berkeley National Laboratory report estimated that data centers used roughly 4 percent of U.S. electricity in 2023 and could use between about 7 and 12 percent by 2028. Power plants and transmission lines take years to build, so new demand runs into a fixed supply. In the PJM grid, which serves much of the Mid-Atlantic and Midwest, the price utilities pay to secure future generating capacity rose roughly ninefold in the 2025–26 auction, with data center demand among the main drivers cited. Utilities pass those costs to customers through regulated rates, with a lag.

Similar bottlenecks have appeared in specialized equipment. Lead times for large transformers and switchgear stretched as utilities and data center developers ordered at the same time. Skilled electrical workers are in short supply in regions with heavy construction.

There is also an indirect channel. AI enthusiasm has lifted technology stocks, and rising wealth tends to support consumer spending. That effect is harder to measure but adds to overall demand.

None of this guarantees broad inflation. Electricity is a small share of the consumer basket, and bottlenecks in one industry can ease as suppliers expand. The concern is persistence: if higher power costs feed into the prices of many goods and services, a narrow shock becomes a broader one.

Monetary Policy is the set of tools the Federal Reserve uses to influence the cost and availability of credit. Its main tool is the federal funds rate, the rate banks charge each other for overnight loans. The Federal Open Market Committee sets a target for that rate at eight scheduled meetings a year, and changes ripple out to the rates on mortgages, car loans, business credit, and savings accounts.

An Interest Rate is the price of borrowing money, expressed as a percentage of the amount borrowed per year. Economists pay close attention to the real interest rate, which subtracts inflation from the stated rate. A loan at 5 percent when inflation runs at 3 percent carries a real cost of about 2 percent. The Fed slows the economy by raising real rates and supports it by lowering them.

The AI cycle presents the Fed with two possible errors. If it treats the current price pressure as temporary and waits for productivity gains that arrive late or not at all, inflation could settle above target. If it raises rates to fight pressures that are about to fade, it could unnecessarily slow the economy. Both mistakes have precedents.

The Fed's tools also fall unevenly. The largest technology firms fund much of their data center spending from operating profits, so higher rates slow them less than they slow home buyers, small businesses, or companies that depend on borrowing. Some firms have begun turning to bond markets and private credit for their largest projects, which could make the sector somewhat more rate-sensitive over time.

One point is often missed. A productivity boom does not automatically mean low interest rates. Faster productivity growth raises the expected return on new investment, increasing demand for capital and pushing up the neutral rate, the level of interest rates that neither stimulates nor restrains the economy. AI could slow inflation while still leaving borrowing costs higher than in the 2010s.

The closest modern parallel is the information technology boom of the 1990s. Businesses spent heavily on computers and networks through the 1980s and early 1990s with little visible effect on productivity; in 1987 the economist Robert Solow observed that the computer age was showing up everywhere except in the productivity statistics. Then, from the mid-1990s, U.S. labor productivity growth roughly doubled compared with the previous two decades.

That episode also tested the Fed. In 1996 and 1997, Chairman Alan Greenspan argued that official data were understating productivity growth and resisted raising rates even as unemployment fell to levels that normally signaled rising inflation. Inflation stayed low, and his bet is widely viewed as having paid off. But the boom ended in the 2001 dot-com bust, a reminder that heavy investment cycles can end abruptly.

Electrification offers a longer view. The economic historian Paul David showed that factories took decades to reorganize around electric motors after the technology became available in the 1880s. The large productivity gains did not appear until the 1920s, once firms redesigned buildings and workflows rather than simply swapping one power source for another.

Both cases support the sequence in this article's title: investment and disruption first, measurable productivity gains later. Neither tells us how long the gap will be this time.

The productivity payoff may be smaller than advocates expect. Economist Daron Acemoglu has estimated that AI could raise total factor productivity by well under 1 percent over a decade, based on the share of tasks it can profitably automate. If the gains are modest, the inflationary first stage would not be followed by much of a second.

The gains could also show up in profits rather than prices. Productivity growth slows inflation only if competition forces firms to pass lower costs on to customers. In industries dominated by a few large firms, cost savings may instead widen margins.

The disinflation could arrive through a bust rather than a boom. If AI revenue fails to justify the spending, companies could cut investment sharply, as they did after 2000. Inflation would fall, but because demand collapsed, not because productivity rose.

Finally, energy constraints could lengthen the first stage. If power supply cannot keep up, electricity costs could stay elevated for years, and data center projects could be delayed, pushing any productivity gains further out.

Other forces may simply matter more. Trade policy, government deficits, housing costs, and labor supply all move inflation, and any of them could swamp the AI effect in either direction.

Several regular data releases will show where the economy is and whether the second stage is arriving.

For now, the AI build-out is a demand shock. It raises the cost of electricity, equipment, and skilled labor before it delivers the productivity that could slow inflation across the economy. History suggests the payoff from a general-purpose technology is real but slow, and that policy mistakes are most likely in the period between investment and payoff.

For the Federal Reserve, the challenge is acting on today's data without dismissing tomorrow's possibilities. For interest rates, the likely result is not a simple path downward. Even a successful AI boom could leave rates higher than in the last decade, because faster growth raises the return on capital. Investment comes first; whether disinflation follows, and how soon, remains an open question.