The U.S. economy is being rebuilt on data centers

The digital gold rush: unprecedented infrastructure

The U.S. economy is undergoing a quiet but massive structural transformation that is redefining land use and national capital flows. According to data analyzed by Epoch AI, spending on data centers, computing hardware, and critical networks reached 0.8% of U.S. Gross Domestic Product (GDP) during the first quarter of the year. This phenomenon transcends the software sphere to consolidate as a physical reality: the country is being redesigned to house the foundations of generative artificial intelligence.

Historically, industrial investment cycles have focused on physical transport or energy infrastructure. However, what we are observing today is a paradigm shift toward 'compute infrastructure.' Unlike the dot-com bubble of the late nineties, where much of the investment was diluted in speculative software, the current era is anchored in tangible installed capacity that requires energy, cooling, and massive industrial land.

The chart that reveals a historic divergence

An analysis by investor and economic expert Steve Rattner, based on official figures from the U.S. Census Bureau, highlights an alarming trend for the traditional economy: while private construction spending in sectors such as retail, housing, and traditional offices is plummeting, data center construction has become the sole growth engine for the sector. Between December 2023 and July 2026, annual investment in this area soared from $24 billion to $75 billion, a net increase of $51 billion. In contrast, the rest of private construction experienced a decline of $120 billion in the same period.

This divergence is so significant that, since 2024, the Census Bureau has begun to break out data centers from the 'office construction' category in its reports. The reason is technical but revealing: the volume of these projects was so disproportionate that it completely distorted the picture of the commercial real estate sector, masking the weakness of conventional workspace construction against the unstoppable rise of physical 'clouds.'

The tip of the iceberg

It is crucial to note that these figures exclusively reflect investment in civil works: the erection of buildings, land preparation, and basic support infrastructure. According to estimates by Cushman & Wakefield, construction accounts for barely 20% of the total cost of a data center. The bulk of the investment—specialized servers, high-performance GPUs, liquid cooling systems, and high-capacity fiber optic networks—remains outside this count. This suggests that the real weight of AI infrastructure in the economy is significantly higher than 0.8% of GDP, placing the United States in a technological arms race with unpredictable consequences.

Consequences and risks of dependency

This concentration of capital raises profound questions about the sustainability and equity of future economic growth:

• Resource displacement: The massive absorption of capital and specialized labor in data center construction is drawing resources away from sectors that traditionally received them. This exacerbates the housing crisis, as capital prefers high-return digital assets over residential brick-and-mortar, complicating access to housing in regions strategic for connectivity.
• The great ROI unknown: There is growing speculation about whether this infrastructure will generate the expected financial return. If AI adoption in the real economy fails to scale its monetization at the same pace as investment, the country could face a severe risk of 'stranded capacity,' a scenario similar to the oversupply of fiber optics after the collapse of the telecommunications bubble in 2000.
• Impact on public services: Local communities are beginning to question the real benefit of these plants. Although the construction phase generates jobs, long-term operations consume massive water and electrical resources—often competing with the residential power grid—in exchange for relatively low local job creation, given that these centers are highly automated.

The U.S. economic dependency on AI infrastructure marks a turning point. We are not just facing a technological evolution, but a shift in the allocation of national capital toward energy-intensive digital assets. The big question for the market is not whether AI is useful, but whether the U.S. economy can bear the weight of such expensive infrastructure without eroding other sectors fundamental to social welfare.