How artificial intelligence is moving the US real economy

Artificial intelligence is no longer just a story about technology, chips, or major Silicon Valley companies. In the United States, the AI ​​boom is permeating the real economy through a wave of investment that spans data centers, electrical grids, construction, employment, and financial markets.

The scale is becoming difficult to ignore. A study by Stijn Van Nieuwerburgh—published by Brookings and also released as a working paper by the National Bureau of Economic Research—estimates that developing infrastructure for AI could require nearly $9 trillion by 2032, assuming 188 gigawatts of new capacity. The Brookings estimate, which incorporates a broader range of investments, reaches $10.3 trillion between 2025 and 2032—equivalent to 3.63% of U.S. GDP annually. While these are projections rather than already committed funds, they help put the scale of the phenomenon into perspective.

The historical comparison is striking. According to Brookings, this projected effort—relative to the size of the economy—would exceed the major U.S. investment cycles seen in canals, railroads, electrification, highways, and telecommunications. The point is not that this technological revolution will unfold exactly like those past transformations, but rather that the infrastructure required to sustain it has reached a macroeconomic scale.

The initial impact is visible in business investment. Major technology companies are allocating extraordinary sums toward servers, data centers, chips, networks, and electrical systems. The Federal Reserve Bank of San Francisco found that spending on information-processing equipment, software, and data center construction accounted for approximately one-third of U.S. business investment in the third quarter of 2025.

This figure helps explain why artificial intelligence has moved beyond being a purely technological bet. When a company builds a data center, the investment ripples out to construction firms, engineers, equipment manufacturers, utility companies, infrastructure providers, and network operators. A portion of the money ends up circulating well beyond the companies that develop the models.

Furthermore, data from the Census Bureau shows that adoption is beginning to spread among firms outside the core technology sector. Between December 2025 and May 2026, approximately 17% to 20% of U.S. businesses were using artificial intelligence. Among companies with 250 or more employees, that figure reached 37%.

This raises an important economic question: if larger companies are the first to capture productivity gains, the new technology could, for a time, widen the gap between large and small firms.

The new economic battleground centers on electricity

The transformation takes on a new dimension when looking at energy consumption. According to an updated report from the Department of Energy and the Lawrence Berkeley National Laboratory, U.S. data centers consumed approximately 176 terawatt-hours (TWh) of electricity in 2023. The same study estimates that, under a central scenario for 2030, consumption could reach 649 TWh—equivalent to 11.8% of the nation’s total electricity use.

The Energy Information Administration (EIA) identifies data centers as a primary driver of the recent surge in electricity demand. This phenomenon is concentrated in specific regions, where operators must integrate large-scale new projects without compromising the stability of the power system.

Virginia serves as an example. The EIA reported that commercial electricity sales in the state rose sharply between 2019 and 2025, driven largely by a high concentration of data centers. In Texas, growing demand is also compelling the grid operator to re-evaluate connection requests and future capacity requirements.

This points to a consequence that could shift the public discourse surrounding artificial intelligence: the technology requires electricity well before it can deliver many of the productivity benefits it promises.

And that pressure has already reached the inflation debate. Researchers linked to the Dallas Federal Reserve estimated that, under certain assumptions, the data center boom could raise wholesale electricity prices and partially pass on to retail prices. Other work from the San Francisco Fed estimates that a moderate data center boom could add between 0.04 and 0.13 percentage points to annual PCE inflation in 2030. If renewable energy expansion were smaller than expected, the effect could be larger.

The inflationary dimension forces us to review one of the first expectations about the new technology. The theory indicates that greater productivity should make it possible to produce more with the same resources and, in this way, alleviate pressures on prices. But before that productivity is realized, infrastructure construction creates enormous demand for capital, electricity, equipment and specialized workers.

The Federal Reserve itself already incorporates that risk. Minutes from their September meeting noted that investments linked to the development of artificial intelligence were contributing to inflationary pressures. Officials also noted that the rise of technological goods associated with infrastructure development had gained importance while the effect of other factors, such as tariffs, moderated.

If the growth in technology investment keeps demand high while supply is slow to respond, the Federal Reserve may face an uncomfortable combination: an economy that grows thanks to investment and, at the same time, inflation that is slow to return to the 2% target.

Productivity, employment and the risk of betting too big

The great economic promise is still ahead. Productivity can become the mechanism that justifies the enormous volume of investment. But the available data advise caution.

The Kansas City Federal Reserve concludes that US productivity has been growing above the pre-pandemic trend, but warns that the acceleration is not evenly distributed across industries. Activities with greater adoption of artificial intelligence show favorable signs, although it is still early to attribute the aggregate increase in productivity exclusively to the new technology.

The labor market offers a similar signal. Projections from the Bureau of Labor Statistics show strong expected growth in data, cybersecurity, software and computing occupations over the next decade. This indicates that the first phase of the process is generating demand for workers capable of building and managing the new infrastructure.

Technology can replace certain tasks and, simultaneously, create new functions and increase demand in complementary sectors. The final result will depend on the speed of adoption, investment and the ability of companies to reorganize their processes.

There is also a financial risk that is beginning to become important. Nieuwerburgh’s study warns that as Big Tech’s spending outstrips its ability to finance it with cash flow, the market is increasingly turning to leasing, joint ventures, project debt, private credit, securitizations and special vehicles.

These structures make it possible to finance a larger volume of infrastructure, but they also distribute the risk over a more extensive chain of investors. The same work points out risks associated with debt, customer concentration, technological obsolescence and delays in project execution.

Therefore, the question is no longer just how much money technology companies are spending. It is who is financing the infrastructure, what profitability it should generate and what will happen if the demand for computing capacity grows less than expected.

If companies build too much, if models generate less revenue than expected, or if the cost of financing increases, some of the investment can become unprofitable. A simultaneous correction in financial markets and capital spending would have much broader consequences than a drop in technology stocks.

The paradox is that technology can be both a source of growth and a pressure factor on the economy. In the short term, the artificial intelligence boom increases demand for electricity, capital, construction and high-tech goods. In the long term, the expectation is that it will allow more to be produced with fewer resources.

The United States is, therefore, building first and harvesting later. The infrastructure is up today; the productivity that must justify it has yet to demonstrate its magnitude.

This mismatch explains why the revolution Technological change is now a matter of economic policy. Artificial intelligence is altering investment decisions, energy demand, financial conditions, and the debate surrounding inflation—even before its full benefits are reflected in productivity statistics.