US Army picks Carlyle and CyrusOne to build AI data centers on its bases

The US Army has launched a scheme unusual for it, sitting at the intersection of defense, infrastructure and AI: instead of a traditional budget-funded build, it is bringing in private capital to create large data centers directly on its bases. The first conditionally selected partners have already been named: Carlyle and CyrusOne.

Who got the projects

The US Army said it has conditionally selected two companies for exclusive negotiations to build and operate commercial hyperscale data centers on military sites. Carlyle won the right to discuss a project at Fort Bliss in Texas, covering roughly 1,384 acres. CyrusOne, which is owned by funds managed by KKR and BlackRock, was chosen for a project at Dugway Proving Ground in Utah, where about 1,201 acres are under consideration for development.

The motive is clear: the military needs ever more computing power for AI-based systems. This is not yet a final contract, only a move into an exclusive negotiation phase. The Army separately stresses that the land is underused but not considered excess, and that the deal is structured through the Enhanced Use Lease program. This legal mechanism lets the military department lease such parcels without losing control over them. In other words, the Army is not selling off territory but trying to turn it into sites for strategic digital infrastructure.

How the scheme works

The key idea is that the private partner carries most of the load. The selected companies must not only invest money in the project but also handle the facility's entire life cycle: from design and construction to operation, maintenance and eventual decommissioning. For the Army, this is a way to bring capacity online faster, without the traditionally long cycle of government construction and without upfront costs for taxpayers. The US Army Corps of Engineers is also involved; it negotiates the leases and provides technical support, including environmental review.

• Private companies finance, build and operate the facilities
• The Army retains ownership of the land
• Initial operation at Fort Bliss is planned for fiscal year 2027
• Dugway Proving Ground is due to reach initial readiness in fiscal year 2029
• Environmental review and agreement on lease terms are required before work begins

The Army also makes an economic argument. The department expects the scheme to bring not only new computing capacity but also significant private investment, jobs and revenue that can be directed to other Army priorities, including quality-of-life infrastructure for service members. The published materials also address the strain on local grids: water and power systems for the sites must be designed to minimize pressure on public utility resources around the bases.

Why the Army needs AI capacity

In the rhetoric of the Pentagon and the Army itself, computing infrastructure is no longer presented as a supporting IT layer but as an element of combat readiness. According to Army Secretary Dan Driscoll, artificial intelligence has become a strategic asset and force multiplier for the Army. In practical terms, this means being able to deploy systems faster for data analysis, autonomous platforms, drone swarms and new command-and-control networks. The closer such capacity sits to the customer and the easier it is to scale, the faster the military can move AI from pilots to working tools.

"AI is a strategic asset for the Army.

These data centers are a critical resource for this task."

It is also telling that the project is tied to a 2025 White House executive order on speeding up federal permitting for data center infrastructure. That means the story is not limited to two sites in Texas and Utah. A broader logic is taking shape at the federal level: if AI is the basis of competitiveness and national security, then it is not enough to build models; the physical infrastructure to run them must be built too. In this scheme the Army acts not merely as a customer but as an anchor partner for major players in the digital and energy industries.

What it means

The news matters beyond the defense sector. It shows that computing capacity for AI is gradually becoming basic infrastructure, like airfields, warehouses or power grids. If the Carlyle and CyrusOne model works, the US Army gets a fast path to new capacity, and the market gets another signal that the next stage of the AI race will play out not only in labs but on the ground, through construction, energy and long-term lease agreements.