The Air Force’s AI Land Rush

This week we honor those who have served. But behind the Veterans Day parades and flags, another story is unfolding — one that could redefine what national security means in the age of AI. The U.S. Department of the Air Force has set a deadline this week for private firms to submit proposals to lease more than 3,000 acres of “underutilized” land on five active military bases (Arnold (Tennessee), Davis-Monthan (Arizona), Edwards (California), Joint Base McGuire-Dix-Lakehurst (New Jersey), and Robins (Georgia)) to build commercial artificial-intelligence data centers—massive facilities each requiring at least 100 megawatts of new power and $500 million in capital investment. Awards are expected early next year. The Air Force says the project will “optimize the real and potential value” of its assets, while helping America stay “at the forefront of innovation.”

At first glance, this looks like an economic-development initiative. In reality, it’s a test case for how far the federal government is willing to go in blending public land, private capital, and national AI strategy.

At first glance, this looks like an economic-development initiative. In reality, it’s a test case for how far the federal government is willing to go in blending public land, private capital, and national AI strategy.

A Familiar Legal Tool, Used in a New Way

The three-week solicitation relies on 10 U.S.C. § 2667, a long-standing law allowing the military to lease underused real estate to non-federal entities in exchange for fair-market value. Known as Enhanced Use Leases (EULs), these agreements have been around for decades. They’ve funded solar farms at Edwards AFB, business parks at Hill AFB, and research labs at Kirtland AFB.

These long-term AI data-center leases, however, represent a sharp departure.

In these earlier cases, the logic was simple: put idle land to work, generate revenue, and build useful infrastructure without new appropriations. The projects—energy, housing, mixed-use—tended to complement military missions.

These long-term AI data-center leases, however, represent a sharp departure. These facilities are not designed primarily for the Department of Defense. The solicitation states explicitly that they are for “private commercial data-center use” and that the government is “under no obligation” to buy services or power from them. The Air Force’s role is essentially that of a landlord.

If we are going to cede public land to private AI companies, why not demand that those companies make AI that is transparent, explainable, effective and democratically governed? Why not demand better pricing for residents and especially for schools? How will communities near these bases be consulted about energy and water use? And who decides when the national-security costs outweigh the economic gains?

Executive Orders and an AI Industrial Policy

This move flows directly from two 2025 executive orders signed by President Donald Trump:

• EO 14179 – Removing Barriers to American Leadership in AI, and

• EO 14318 – Accelerating Federal Permitting of Data Center Infrastructure.

Together they direct federal agencies to identify public land for data-center development and streamline environmental reviews. The Air Force solicitation explicitly cites both orders, calling the proposed projects “Qualifying Projects” under EO 14318.

In other words, what began as a real-estate management tool has become an instrument of national industrial policy, designed to expand U.S. AI computing capacity through private investment on federal property.

Security and Governance Questions

Experts are uneasy. Stacie Pettyjohn of the Center for a New American Security told Defense One she had “never heard of something like this before.” Leasing thousands of acres inside active installations—particularly at Edwards AFB, home to sensitive flight-testing programs—could invite espionage risks and make it harder for the military to reclaim control later.

If executed as written, the leases could last 50 years or longer. Developers would be responsible for their own power and water, and the Air Force would hold only a first right of refusal if a power plant is built on site. In a moment when data sovereignty and critical-infrastructure security dominate public debate, granting private companies long-term control over land within defense perimeters is, at minimum, a provocative experiment.

Why It Matters

America’s AI ambitions are colliding with real constraints—land, power, and permitting delays. Using military bases as data-center real estate is a clever workaround. Yet it raises foundational questions about what counts as a public asset and how national strategy should be governed.

Without even addressing the potential for sweetheart deals, this land deal misses the opportunity to make any demands of these companies. If we are going to cede public land to private AI companies, why not demand that those companies make AI that is transparent, explainable, effective and democratically governed? Why not demand better pricing for residents and especially for schools? How will communities near these bases be consulted about energy and water use? And who decides when the national-security costs outweigh the economic gains?

These are not questions of procurement minutiae. They go to the heart of how a democracy manages the interface between government, technology, and private capital in the age of artificial intelligence.

Serving the Public in the AI Era

Veterans Day is a reminder that military service has always involved more than firepower; it’s about stewarding the public good. The Air Force’s new leasing plan asks whether the same can be said of public-private AI partnerships.

Those who serve take an oath to defend the Constitution, not a market. 

The Air Force’s solicitation is legal, even routine, under the Enhanced Use Lease authority. But its purpose is unprecedented. The same statute that once enabled a solar array now serves to launch a new chapter in America’s AI build-out—inside the fences of its military installations.

How this experiment unfolds will tell us a great deal about the future of public-private governance in the AI era: who owns the infrastructure, who bears the risks, and who gets to decide.

One thing is very clear: the one deciding isn't you or I.