Meta is stuffing AI chips into tents: building data centers in half a year—“6 times faster than traditional methods”—and borrowing playbooks from Tesla and xAI.

Meta has built six industrial tents in New Albany, Ohio, filled with AI chips worth billions of dollars, replicating the emergency tactics Tesla used when it rushed to produce the Model 3. Traditional construction takes 2-3 years, while the tents can be completed in 4 to 6 months.
(Background: AI stock guru Serenity: The EU Chips Act 2.0 includes “silicon photonics,” delivering long-term upside for the AI data center supply chain)
(Additional context: Put data centers in your home! SPAN calls for using 80,000 nodes to build a decentralized AI computing network)

Cleanview, a market intelligence analysis firm focused on U.S. power infrastructure and data center development, has recently tracked Meta’s construction permits and satellite imagery for the project in New Albany, Ohio. It found that the six massive tents had transformed from vacant lots into computing facilities within a few months. Inside are AI chips worth billions of dollars, operating 24 hours a day.

To shorten the time to bring computing capacity online, Meta has discarded all the baggage of traditional building regulations, letting industrial tents replace permanent structures as the front line of AI infrastructure.

Tesla’s old tricks, Meta’s new battlefield

After investigating local building permits, Michael Thomas found that from April to June this year, Meta applied for construction permits for five 125,000-square-foot (about 11,600 ping) tents in stages. The satellite images he shared on X show that all the tents have been fully completed—so fast it’s astonishing.

Meta is building dozens of massive tents at campuses across the US, sticking billions of dollars of chips inside, and powering them with off-grid turbines.

The AI race has officially entered its Mad Max phase.

Over the last month, I reviewed hundreds of documents and satellite… pic.twitter.com/U8yDZUlEO0

— Michael Thomas (@curious_founder) June 4, 2026

How long does traditional construction take? Meta’s first five permanent data center building projects in New Albany each took 2 to 3 years to be completed. The tents compress this timeline to 4 to 6 months—cutting the time by more than half.

The inspiration for this “emergency deployment structure” isn’t mysterious. Last year, Zuckerberg told The Information that he planned to house Meta’s multi-Gigawatt data centers using weatherproof tents as a way to accelerate expansion. At the time, most in the industry were half-believing and half-doubting—now, the satellite images speak for themselves.

The precedent for the tent strategy was Tesla’s 2018 production-line tents set up in the parking lot of the Fremont factory in California. It was an emergency measure to rush Model 3 deliveries at the time—Musk himself slept on the factory floor to oversee the work. Now, the same logic has been transferred to the AI computing power race: speed comes first, and permanent facilities are a luxury.

The moat of computing power—energy determines the ceiling

Tents are just containers. The real question is: where does the power come from?

The power-supply installations for this batch of tents in New Albany are code-named Socrates. They use a series of modular gas turbines and fast-start reciprocating engines, with a total power capacity of 200 Megawatts. The main equipment includes Solar Turbines Titan 250 gas turbines, Siemens Energy SGT400, and 15 Caterpillar 3520 series reciprocating engines.

With modular gas turbines powering the system, the plan was pioneered by Musk’s xAI. In simple terms, connecting to the traditional power grid requires waiting for the utility’s scheduling and for new lines to be pulled in, often taking years; modular gas generator sets can be installed on site within months. xAI’s Colossus supercomputing cluster in Memphis, Tennessee, is powered using this setup for rapid power—Meta has fully placed its order and adopted it.

This also means Meta’s definition of “data center” has quietly expanded. It’s no longer just permanent brick-and-mortar buildings, but any kind of container that can be deployed quickly, powered on quickly, and run chips—whether that container is made of steel, concrete, or industrial canvas.