Meta has started housing AI data centers in tents. Not a metaphor. Actual weatherproof structures stuffed with chips, planted in a field outside New Albany, Ohio. According to TechCrunch AI, the company has put up six of these “rapid deployment structures” to cut data center construction time roughly in half.
The findings come from Michael Thomas, founder of Cleanview, a firm that tracks data center deployments. He reviewed local city permits and shared satellite images on X. TechCrunch AI reports that Meta started building five 125,000-square-foot tents between April and June, and the satellite shots show they’re already standing.
What’s actually happening
This isn’t a brand-new plan. Mark Zuckerberg told The Information last year he wanted to use weatherproof tents to house Meta’s multi-gigawatt data centers. What’s new is the proof of speed and scale. Permits plus satellite imagery show how fast this went from idea to built.
A few specifics worth pinning down:
- Six tents, five of them 125,000 square feet each.
- Built in roughly three months, April through June.
- Powered by 200 megawatts of modular gas turbines sitting nearby.
- Inside: AI chips that are, in Thomas’s estimate, likely worth billions.
Where the playbook comes from
What stands out here is that Meta didn’t invent this. It borrowed.
The tents echo what Tesla did in 2018, when it threw up a structure in the parking lot of its Fremont factory to rush Model 3 production out the door. Elon Musk called it a “general assembly line” under a tent. It looked desperate at the time. It worked.
The on-site gas turbines come from xAI’s playbook. Musk’s AI company popularized wheeling in modular gas turbines to power compute fast, skipping the years-long wait for grid connections and permanent substations. Meta is now running the same move.
So you’ve got one company copying two of Musk’s speed hacks at once. That tells you something about the pressure everyone’s under.
Why this matters
The normal way to build a data center is slow and expensive. Permanent buildings, grid hookups, environmental reviews, the works. That can take years. In an AI race measured in months, years is a death sentence.
Tents flip the math. You trade polish for speed. The chips inside don’t care whether the roof is steel or fabric, as long as they stay cooled and powered. If you can stand up gigawatts of compute in a quarter instead of waiting on construction crews and utility timelines, you ship models sooner.
There’s also a cost angle. Meta has said it plans to spend up to $145 billion on data centers and other capital expenditures. Wall Street has not loved that number. The stock is down about 5% this year, per TechCrunch AI. Putting chips in tents is one blunt way to trim the bill and show investors you’re not torching cash on concrete.
The awkward backdrop
Here’s the tension. All this infrastructure is going up while Meta struggles to actually ship its AI. A recent Wall Street Journal report cited by TechCrunch AI found that Meta’s latest model, Muse Spark, is finished, but the developer APIs needed to access it have been delayed again and again.
Think about that. Billions of chips humming inside tents, and developers still can’t get their hands on the model those chips are meant to serve. The compute is racing ahead of the product.
What to watch next
A few things I’d keep an eye on:
- Does the tent strategy spread? If Meta proves fabric-and-turbine sites work at scale, expect rivals to copy the copiers.
- Muse Spark’s API launch. The hardware story only pays off when developers can build on the model. Watch for that release.
- Grid and emissions pushback. Gas turbines parked next to tents will draw scrutiny from regulators and local communities. Speed has a cost.
Meta hasn’t commented yet. TechCrunch AI says it has reached out and will update if the company responds. For now, the picture is clear enough: the AI buildout has gotten so frantic that the biggest players would rather put billions of dollars of silicon under a tent than wait. Full details are at the original source.