Air Force officials want private companies to build artificial intelligence data centers on more than 3,000 acres of land on five of its military bases, raising questions about security, ethics, and land use.
The Air Force bases—Davis-Monthan in Ariz., Edwards in Calif., Robins in Ga., Joint Base McGuire-Dix-Lakehurst in N.J., and Arnold in Tenn.—will make “underutilized” land available “for private commercial data center use,” according to an Oct. 15 and Oct. 21 request for lease proposal from the service published online.
“AI is transforming the modern world, and these data centers are crucial for America to remain at the forefront of innovation,” Robert Moriarty, the Air Force’s deputy assistant secretary of installations, said in an unpublished news release provided to Defense One.
The lease proposal follows a late July executive order in which President Donald Trump promised a “golden age for American manufacturing and technological dominance” by offering up federal land for AI data centers. But a national security expert said the unprecedented deal could blur the lines between public and private partnerships—and may make it hard for the military to regain control of that land in the future.
“I have never heard of something like this before, where some of the public land was going to be leased to private companies to use,” said Stacie Pettyjohn, a senior fellow and director of the Defense Program at the Center for a New American Security think tank. “I think it is noteworthy … because it is potentially ceding land that the U.S. government will actually never get control over again.”
The vast majority of the land, 2,115 acres, is spread across seven sites at Edwards Air Force Base, home to many of the service’s test aircraft projects. Pettyjohn said a private company having access to those bases and nearby training exercises and test facilities is concerning.
“If you’re letting a lot of people on the base, there is a lot of room for potential espionage, let alone sabotage or things like that,” Pettyjohn said. “It does seem to be purely commercial and transactional, and a part of the broader administration’s policy towards blurring the lines between private and public which, in some ways, is what you see in authoritarian states like China.”
Spaces at the other bases in the proposal include 300 acres in Arizona, 219 acres in Georgia., 193 acres in New Jersey, and 274 in Tennessee. An Air Force spokesperson said the service will “ensure the proper security precautions are taken to protect the installation and its assets.”
The AI boom has driven a need for land as well as for more electrical power to fuel the data centers, underscoring the Pentagon’s rapid search for companies to field nuclear microreactors as a supply source.
For the Air Force’s proposal, the chosen developer would be responsible for supplying “sufficient power and water,” with interest given to water-efficient technology, the service spokesperson said.
Additionally, companies must submit a “mitigation and contingency plan to ensure the local communities’ utilities service life, resiliency, and capacity are not impacted,” the spokesperson said.
The government can purchase AI data center services and power from the project, but “is under no obligation to do so,” according to the lease proposal. Companies must offer “fair-market value” for the land and a payment of $250,000 to the government. The developer must also have a prior history and show proof they’ve built three AI data centers drawing at least 100 megawatts of new contracted power within the last three years.
Responses to the Air Force’s solicitation are due by Nov. 14 and the service will select winning lease proposals by January 2026.
“The Department of the Air Force is focused on swiftly, yet effectively, selecting an industry leader as a lease applicant,” said Benjamin Kindt, the Air Force’s chief of real estate development, in the unpublished news release.
The post The Air Force wants to put private AI data centers on its bases, raising security, land-use fears appeared first on Defense One.
