America’s next wave of artificial intelligence infrastructure may rise not only in Silicon Valley, Northern Virginia, or Texas, but on land entrusted to the military.

In March, the Army conditionally selected two private companies for separate commercial hyperscale data center projects. Carlyle, a Washington, D.C.-based investment firm, was selected for the Fort Bliss project in Texas. CyrusOne, a Dallas-based data center operator backed by funds managed by KKR and BlackRock, was selected for Dugway Proving Ground, Utah. The companies would finance, construct, operate, and eventually decommission the facilities on Army land considered underused, though not surplus. The Army says the projects will expand computing capacity for the warfighter and improve installation resilience, though the final lease terms — and the precise military benefits — remain under negotiation.

The Air Force is pursuing a similar course at multiple installations. Its own public explanation, however, exposes the policy question the Pentagon has not adequately answered.

The Air Force acknowledges that the proposed facilities will be privately owned and “will primarily support commercial use.” More strikingly, the service is neither requiring nor guaranteeing Air Force access to its computing capacity. In return for the land, the service expects cash or in-kind consideration. Where a developer builds dedicated power generation, the Air Force would receive a first right of refusal during a grid outage — but any actual purchase of that power would require another agreement.

That should stop Congress and the Pentagon long enough to ask a basic question: Why is the military placing privately owned, primarily commercial data centers on its installations, and what enforceable military benefit will the nation receive in return? The answer should be straightforward: No commercial data center should occupy military land unless the agreement provides contractually guaranteed benefits to military readiness.

The United States unquestionably needs more computing capacity. Artificial intelligence is already being applied across command and control, operational planning, logistics, intelligence, cyber operations, weapons testing, and autonomous systems. America cannot compete with China or prepare for future warfare without the chips, servers, electricity, cooling, communications, and secure facilities that make AI possible. Compute is already a strategic resource.

Military installations may offer significant advantages. They provide large parcels of controlled federal land, proximity to communications and security infrastructure, and potential room for dedicated power generation. Private financing may also deliver capacity faster than traditional military construction and reduce immediate taxpayer costs.

But military land is not simply vacant real estate, and the Pentagon is not an ordinary landlord.

Federal law permits military departments to lease non-excess property when doing so promotes national defense or serves the public interest. It also generally requires cash or in-kind consideration worth at least the fair market value of the lease. Those are important protections, but fair market rent is not the same as military readiness.

A commercial data center does not become a defense capability merely because it sits on military property. Proximity is not access. Commercial capacity is not assured military capacity. And lease revenue is not a substitute for operational advantage.

The Army’s announcement says its projects will enhance computing capability for the warfighter. That is encouraging. Yet the public documents do not explain how much computing capacity the Army will receive, what it will cost, whether access will be guaranteed during war, or whether the government will take priority over commercial customers. Those questions may be resolved in the final agreements. They must be.

The Air Force proposal is even more troubling because the service expressly disclaims any guaranteed access to the data centers’ computing capacity. A first right of refusal for emergency power is useful, but weaker than assured access. And if Air Force organizations must later negotiate separately to purchase computing services from a facility built on Air Force land, the national security justification begins to look thin.

At minimum, every such lease should guarantee several benefits.

The military should receive a defined share of computing capacity at pre-negotiated rates, with priority access during war, mobilization, cyberattack, or national emergency and government step-in rights if the operator cannot perform. That capacity must meet the security and accreditation requirements for classified or mission-sensitive workloads. The facilities should add enough generation, storage, cooling, and water capacity to avoid competing with installation missions or burdening nearby communities. Ownership, financing, hardware, software, contractors, and tenants should be protected from hostile foreign control and compromised supply chains.

The agreements should also establish rigorous physical and cyber separation between military systems and commercial operations; government inspection and audit rights; tested continuity plans; and clear provisions for ownership changes, bankruptcy, mission expansion, and eventual decommissioning. If a hyperscale data center increases an installation’s value as a target for espionage, sabotage, cyberattack, or long-range strike, that risk must be assessed rather than ignored.

The administration has already articulated the right principle. National Security Presidential Memorandum 11, issued in June 2026, directs the government to use contractual clauses or other means to ensure that no commercial entity can prevent, disable, degrade, or secretly modify an AI system on which American warfighters depend. It also calls for a roadmap to provide the national security enterprise adequate access to advanced computing resources and high-security AI facilities.

That standard should apply not only to AI models but to the infrastructure beneath them. The nation should never discover during a crisis that a private company controls the computing power, energy, maintenance, or permissions required for a military mission.

Scripture offers a caution about responsibility that cannot safely be outsourced. Jesus contrasted the shepherd who owns the flock with the hired hand who abandons it when danger comes: “He who is a hired hand and not a shepherd, who does not own the sheep, sees the wolf coming and leaves the sheep and flees, and the wolf snatches them and scatters them. He flees because he is a hired hand and cares nothing for the sheep” (John 10:12-13, ESV). The lesson is not that every contractor is faithless, but that government cannot surrender its ultimate responsibility for the mission or the lives of service members. A commercial operator answers to investors first, and its incentives can shift when the agreement grows costly or dangerous.

Private enterprise is not the enemy. Properly structured, these partnerships could strengthen the military, accelerate American AI development, improve installation resilience, and benefit surrounding communities. Private profit may accompany the national defense mission.

It cannot replace that mission.

America needs more compute. But before the Pentagon becomes Big Tech’s landlord, it must show what the warfighter receives in peace, crisis, mobilization, and war — and guarantee those benefits contractually.