It’s easy to tell that the AI revolution is underway, but there’s nothing better to see it in black and white in data.
Data of US construction activity shows that since late 2022 when ChatGPT was released, datacenter construction has accelerated at the expense of construction of office space and retail buildings. According to U.S. Census Bureau data, data center construction has soared to approximately $45 billion annually as of early 2025, up from roughly $15 billion in late 2022. This represents a threefold increase in just over two years, marking the steepest climb in the sector’s history.
Meanwhile, office construction has plummeted to around $50 billion annually, its lowest level since the early 2010s and down from peaks above $120 billion in the early 2000s. The divergence between these two categories has never been more pronounced, with data centers now approaching parity with office construction for the first time since records began in 1995.
The timing is striking. ChatGPT’s November 2022 launch catalyzed an AI arms race among tech giants, each racing to build the massive computational infrastructure required to train and deploy large language models. These facilities demand enormous power supplies, sophisticated cooling systems, and specialized networking equipment—driving construction costs far beyond traditional commercial projects.
Major technology companies have announced tens of billions in data center investments. The construction boom spans traditional tech hubs and newer markets, with developers racing to secure power capacity and suitable land. Some projects involve facilities exceeding one million square feet, consuming as much electricity as small cities.
The office sector’s decline reflects a different revolution: the lasting impact of remote work. Even as some companies push return-to-office mandates, construction data suggests developers remain skeptical about long-term demand. Office construction has been falling steadily since 2020, but the pace has accelerated recently, dropping nearly 40% from 2023 levels. The fall coincides with many companies laying off large numbers of employees and replacing them with AI.
This construction shift represents more than a temporary trend. It signals a fundamental reordering of American commercial real estate priorities, where the infrastructure for training AI models now rivals—and may soon surpass—the physical spaces where knowledge workers traditionally gathered. Whether this represents a permanent transformation or a speculative bubble remains to be seen, but the current trajectory is unmistakable: artificial intelligence is literally reshaping the American built environment.