Data Center Shell Construction Hits $75 Billion Annualized Rate, Up 60% Year-Over-Year

To the technology sector and its financial backers, the unprecedented $75 billion annualized pace of data center construction in July represents the essential physical foundation for an artificial intelligence economy that will define the next decade. To local municipalities and utility regulators, that same exponential building curve is an immediate threat to regional power grids and water supplies, prompting a wave of legislative pushback designed to halt the concrete pouring before infrastructure fails. The tension between hyperscale computing demands and local resource limits has become the defining conflict in commercial real estate for 2026.[2][3]

Data released by the U.S. Census Bureau revealed that spending on AI data center "shell" construction surged to a seasonally adjusted annual rate of more than $75 billion in July. That figure represents a nearly 60% increase from July 2025 levels. Crucially, this metric captures only the physical building costs—labor, materials, contractor profits, and engineering work for the warehouse-scale structures—and entirely excludes the far larger capital expenditures for the servers, processors, memory chips, and fiber-optic connections that eventually go inside.[2][3]

The data center boom is currently masking a significant contraction across the rest of the commercial real estate and construction sectors. Total U.S. construction spending actually fell 0.5% in July to an adjusted annual rate of $2.157 trillion, missing market expectations of a flat reading and hitting its lowest overall level since October 2023. Private nonresidential construction managed a slight 0.4% gain for the month, but the underlying mechanics of that growth are entirely dependent on digital infrastructure.[1][4]

The divergence between AI facilities and traditional commercial real estate has never been wider. "The increase in nonresidential construction spending that occurred in July was entirely due to data centers," stated Associated Builders and Contractors Chief Economist Anirban Basu in the firm's analysis of the July data. "Excluding that booming category, nonresidential spending fell for the second straight month and is down to the lowest level since September 2023."[4]

While data center shells multiply, other major categories are retreating rapidly. Manufacturing construction declined nearly 22% year-over-year in July as the initial stimulus from the CHIPS and Science Act began to fade. During the first seven months of 2026, total construction spending reached $1.244 trillion, which remains down 3.5% compared to the same period in 2025. The shift illustrates a commercial real estate market where capital is aggressively rotating out of traditional industrial and office assets and into specialized, power-dense digital infrastructure.[1]

As the physical footprint expands, the political and logistical friction is escalating. The acceleration is happening even as political backlash against data centers grows in the communities where they are sited. Local jurisdictions are increasingly citing concerns over soaring electricity costs, potential blackouts, and the localized strain on water resources required for advanced cooling systems.[2]

For the commercial real estate lenders and private equity firms financing this build-out, the gap between shell construction and energized capacity represents a growing vulnerability. The $75 billion annualized pace measures capital going into the ground, not computing capacity coming online. Supply chain bottlenecks for specialized transformers and high-voltage switchgear mean that a completed physical shell may sit empty for months before it can be connected to the regional grid and begin generating lease revenue.[2]

The trajectory of this asset class now depends entirely on utility interconnection timelines rather than construction capacity. With monthly construction spending on data centers having spiked exponentially since the beginning of 2021, the industry is testing the physical limits of the U.S. electrical grid. The next phase of the AI real estate cycle will be defined not by how fast developers can pour concrete, but by how quickly they can secure the gigawatts of power required to turn those $75 billion shells into functioning data centers.[2][4]