Meta's Tulsa data center pairs 1,500 MW of clean energy with closed-loop liquid cooling

A single site with fleet-scale numbers

Meta says the Tulsa facility is its first in Oklahoma, its 28th in the US, and its 32nd globally. The framing matters: this is not an experiment but the incremental addition to a mature, standardized fleet.

The economics are lopsided in a way worth naming plainly. Meta projects over 1,000 construction workers at peak but only about 100 permanent operational jobs once the site is running. That gap is structural to AI-optimized data centers — they are capital-dense and labor-light by design. The $1 billion-plus investment buys compute capacity and physical plant, not a large ongoing payroll.

The workforce development partnership with Tulsa Tech and Tulsa Community College, aimed at 200+ graduates annually in trades like cooling simulation, fiber optics, and structured cabling, is Meta's answer to that gap — training a regional talent pool that can serve infrastructure work beyond this one building.

Cooling design as the load-bearing claim

The most technically specific commitment is the cooling architecture. Meta says Tulsa will use a closed-loop, liquid-cooled system that recirculates the same water and will use zero water for a majority of the year.

This is the pointed detail because water is the pressure point for AI data centers in many regions. A closed-loop liquid system is also the cooling approach that dense AI accelerators increasingly require — so the water-efficiency story and the AI-workload story are the same engineering decision, not two separate initiatives.

Meta reinforces it with offsets: paying full water and wastewater service costs so they aren't passed to consumers, a fund with the Tulsa Community Foundation for residents' water bills, and a ten-year Phytech project deploying plant-sensor irrigation across roughly 1,500 acres of corn, soybeans, and winter wheat to save over 50 million gallons per year.

The grid as part of the deal

On energy, Meta commits to matching the data center's electricity use with 100% clean energy and adding over 1,500 megawatts of clean energy to the Oklahoma grid. That megawatt figure is the real measure of the site's scale — it is the demand side of the compute the building will house.

Through our utility bills, we'll be spending hundreds of millions of dollars on grid infrastructure, like substations and transmission lines, that benefit all consumers and the data center's electricity use will be matched with 100% clean energy.Montana Labs

The company also commits to paying the full cost of its energy use and contributing to PSO's Light a Life bill-assistance program. The recurring pattern across water and power is the same: Meta pre-empts the cost-shifting objection by pledging to absorb infrastructure costs itself.

What the Tulsa template says about AI capacity planning

Read as an engineering document rather than a press release, this announcement shows that siting AI compute is now a resource-negotiation problem as much as a hardware one. Every headline commitment — closed-loop cooling, 1,500 MW of clean energy, full cost coverage — addresses a specific local constraint that could otherwise slow or block the build.

For teams planning their own AI infrastructure, the lesson is that compute capacity is increasingly gated by water and power availability and by the social license to consume them. Meta's Tulsa design treats those constraints as first-order requirements, engineered into the cooling loop and the grid contributions rather than mitigated after the fact.