Meta turns to tent-based data centers to accelerate AI infrastructure

Meta has begun using “rapid deployment structures”—essentially 125,000-square-foot tents—in New Albany, Ohio, to house its data centers. The company is deploying these structures in an effort to cut traditional data center construction time in half. Michael Thomas, the founder of Cleanview, a firm that tracks data center deployments, tracked the project’s progress. According to local permits, Meta began building five of the tents between April and June, and satellite images show that six tents have now been constructed at the Ohio site. While the physical deployments are now visible, the strategy itself is not entirely new. Meta CEO Mark Zuckerberg spoke to The Information last year about his plan to use weatherproof tents to house the company’s multi-gigawatt data centers. However, Thomas’ review of local permits and satellite images showcases the speed of construction and the massive scale of the project.

The strategy borrows tactics from Tesla’s factory construction, reminiscent of the tents Tesla built in the parking lot of its Fremont, California factory when it was rushing to roll out its vehicles. To support the power requirements of the Ohio facility, the site is powered by 200 megawatts of modular gas turbines, a tactic popularized by competitor xAI. Inside these rapid deployment structures, AI chips, which are likely worth billions of dollars, will operate.

This physical infrastructure push comes as Meta faces pressure to manage its massive capital expenditures—defined as the funds a company uses to acquire, upgrade, and maintain physical assets. Meta intends to spend up to $145 billion on data centers and other capital expenditures. Wall Street has reacted coolly to the scale of this spending, with Meta’s stock trading down 5% this year. Putting AI chips in tents represents one way the company is attempting to trim its overall infrastructure bill.

At the same time, Meta is navigating delays in its software pipeline. The Wall Street Journal reported that Meta’s latest AI model, Muse Spark, is complete, but the APIs that developers rely on to access the model have been repeatedly delayed.

Why it matters

Meta’s shift to rapid deployment structures highlights the extreme measures tech giants are taking to bypass traditional construction bottlenecks while managing massive capital expenditure requirements for AI.