We shape our buildings and afterwards our buildings shape us. – Winston Churchill
We shape our buildings and afterwards our buildings shape us.
U.S. commercial real estate has reached an unusual milestone: The country is now spending more to build data centers than offices. Data center construction has surged to a seasonally adjusted annual rate of $49.5 billion, surpassing office construction at $43.4 billion. Offices once represented the physical center of the white-collar economy, but office spending started to plateau during COVID-19. Before ChatGPT launched in late 2022, office construction spending exceeded data center construction by more than $50 billion annually. Data centers now command more construction dollars, reflecting how quickly capital has shifted toward AI infrastructure and the demand for computational power.
Large data centers are labor-intensive to build, requiring electricians, plumbers, engineers, construction workers, project managers and specialized contractors before a facility opens. The Wall Street Journal reported in February 2025 that about 1,500 people were building the first Stargate AI data center in Abilene, Texas, while the completed facility was expected to have about 100 full-time employees. Jim Grice, a real estate and project finance attorney focused on data centers, put it clearly: “Data centers are very labor-intensive to build, not as labor-intensive to operate.”
The labor story changes once a facility opens. Data centers and offices carry very different labor footprints. A 250,000-square-foot data center might have approximately 50 full-time workers, according to a Virginia legislative review, equal to roughly one worker per 5,000 square feet. Common office-space standards recommend 125 to 225 square feet per person, or roughly 1,100 to 2,000 workers in the same amount of space. Put another way, data centers are about 20 times less labor-intensive by square footage than office space.
A new office building can bring thousands of workers into a neighborhood each day, supporting restaurants, transit, retail and other local businesses. A data center leaves its mark through power demand, electrical equipment, cooling requirements, land use, local tax revenue and grid investment. Development also depends on land, permitting, transmission, substations, backup generation and interconnection timelines, all of which can shape the speed and cost of a buildout. In the Lake Tahoe region that straddles Northern California and Nevada, Liberty Utilities is being forced to find a new supplier for roughly 49,000 California residents as NV Energy, its major current supplier, redirects power toward Nevada’s growing data center demand. The episode shows how once abstract electricity forecasts are turning into utility decisions that affect residential customers.
The crossover between data centers and offices underscores how much of the AI buildout has moved into the physical economy. Peter Orszag, CEO of financial firm Lazard, recently described the U.S. economy as a “levered bet on AI,” pointing to artificial intelligence and high-income consumers as key sources of growth. That dependence is now visible beyond equity markets. Data centers show where part of that leverage is settling: in construction spending, electricity demand, infrastructure capacity and local power systems being asked to absorb a new kind of commercial real estate cycle.
Between the Lines is a weekly blog by DoubleLine Portfolio Managers Sam Garza, Joseph Mezyk and Quant Analysts Fei He, CFA and Sunyu Wang that breaks down topical macro and market issues. For questions or suggestions please e-mail us at betweenthelines@doubleline.com. The views and opinions expressed herein are those of the authors and do not necessarily reflect the views of DoubleLine Capital LP, its affiliates or employees.
