The latest U.S. construction data illustrate a major shift in where corporate capital is being deployed. Data center construction has surged to an annualized rate of roughly $85 billion, driven by the rapid expansion of artificial intelligence and cloud infrastructure, while traditional office construction remains under pressure from structural changes in workplace demand.

The divergence is occurring within a broader construction market that remains mixed. Total U.S. construction spending increased 0.9% in August to a seasonally adjusted annual rate of $2.203 trillion, although it remained 1.7% below August 2025 levels. Private nonresidential construction rose 1.0% from July, making the composition of spending increasingly important to understanding the underlying investment cycle.

AI Infrastructure Has Become a Major Construction Engine

The scale of the data center increase is significant. According to the latest figures shown in the attached data, spending reached approximately $85 billion on an annualized basis in August, representing a 73% year-over-year increase. KPMG’s analysis of the Census Bureau data similarly reported that data center construction spending reached a record $85 billion annualized rate and increased 73.2% from a year earlier.

The increase reflects more than the construction of server buildings themselves. AI infrastructure requires large investments in power systems, cooling equipment, electrical components, networking infrastructure and specialized industrial equipment. As hyperscalers and other technology companies expand computing capacity, the resulting capital expenditure is increasingly flowing through the broader construction and industrial supply chain.

The growth is particularly notable because it has occurred at a pace far beyond the traditional construction cycle. Since the beginning of 2021, the attached figures show annualized data center construction spending increasing by approximately $76 billion, or 823%. This suggests that AI is not simply generating demand for additional computing hardware but is influencing the physical allocation of capital across the U.S. economy.

The Office Construction Market Is Moving in the Opposite Direction

The contrast with traditional office construction is equally important. General office construction spending fell to approximately $46 billion on an annualized basis in August, according to the attached data. The figure is approximately 10% below the previous year and represents one of the lowest levels in the series since 2015.

The divergence reflects a fundamental change in the economics of commercial real estate. Hybrid and remote work have reduced demand for conventional office space in many markets, while high vacancy rates and uncertainty over future workplace requirements have constrained new development. As a result, capital that once supported conventional office construction is increasingly being directed toward infrastructure associated with computing, energy and digital services.

The result is a record $39 billion gap between annualized data center and general office construction spending. The gap is significant because it illustrates how technological change can alter the physical economy: buildings designed to support artificial intelligence are attracting substantially more incremental construction capital than buildings designed primarily for traditional office employment.

The AI Investment Cycle Extends Beyond Technology Companies

The construction figures also show why the AI investment cycle should not be viewed solely through the performance of semiconductor and software companies. Data centers require large amounts of electricity, specialized cooling systems, generators, transformers, construction machinery and other physical infrastructure. The spending boom therefore creates potential demand across a much wider group of industrial and infrastructure suppliers.

Energy infrastructure is becoming particularly important as developers encounter limitations in grid connections. Reuters recently reported that demand for smaller gas turbines is increasing as data center developers seek faster sources of behind-the-meter electricity while waiting for permanent grid capacity. The research cited by Reuters projects nearly 30 gigawatts of behind-the-meter gas generation could be added by 2030, with data centers accounting for the vast majority.

This creates an important distinction between AI investment and AI-related construction. The first involves computing equipment and software, while the second encompasses the physical infrastructure required to operate those systems. Together, they represent a much broader capital-spending cycle with implications for construction, utilities, industrial equipment and energy markets.

Going forward, the key question is whether the current pace of data center investment can remain economically justified as spending reaches unprecedented levels. The market will need to monitor AI demand, data center utilization, electricity availability, financing costs and the returns generated by increasingly large infrastructure investments. For the U.S. economy, the widening gap with traditional office construction already demonstrates that AI is changing not only the technology sector but also the allocation of physical capital across the broader economy.