Pentagon Advances Construction of Large-Scale AI Data Centers at Over Ten U.S. Military Bases
According to information from military and media sources, the Pentagon is advancing a plan for large-scale AI data centers that combine commercial and military uses at over ten military bases across the U.S. through the Army and Air Force. This includes bidding for large areas of military land at bases such as Fort Bliss in Texas, Dugway Proving Ground in Utah, and Fort Hood and Fort Bragg.
The Army has conditionally selected Carlyle Group to develop an approximately 1,384-acre AI data center park at Fort Bliss, while CyrusOne, jointly owned by KKR and BlackRock, will build similar facilities at Dugway Proving Ground, with each project estimated to require an investment of about $2 billion. The Air Force is also soliciting private companies to build AI data centers on "underutilized but not excess" land at several bases in Tennessee, California, New Jersey, Arizona, Georgia, and Alaska.
In terms of market mechanisms, the Pentagon is using an enhanced "enhanced use lease" framework to allow private companies to finance, build, operate, and eventually decommission long-term military land without charging upfront rent. The military exchanges land and security environments for access to large-scale computing power and AI infrastructure; projects must generate their own power and achieve "net zero water use," and cannot directly connect to local power grids. This has led to scrutiny and backlash from Congress and local communities regarding energy consumption, water resources, and potential military target risks.
Source: Public Information
ABAB AI Insight
Historically, the U.S. military has relied on partnerships with large cloud service providers for information infrastructure, such as various cloud and AI projects between DoD and AWS, Azure, and Google Cloud. The construction of large-scale data centers driven by private capital on military land marks a structural shift from "using commercial clouds" to "building sovereign computing," embedding AI training and inference capabilities directly into military-controlled physical spaces and locking in infrastructure layouts with long-term leases.
In terms of capital pathways, this model outsources the capital expenditure for AI infrastructure to Wall Street and specialized data center operators: companies like Carlyle and CyrusOne will bear billions of dollars in construction and operational costs in exchange for long-term lease revenues and commercial rights to provide hosting services for AI companies and the military. The military uses military base land and security environments as "contributions," while computing power usage rights and strategic dominance serve as "returns," forming a closed loop of "land + security environment → capital expenditure → AI infrastructure → military computing," turning military AI infrastructure into long-term assets for private equity and infrastructure funds.
In terms of analogy and industry positioning, this combination of "military bases + private data centers" can be compared to the construction of base facilities by military contractors during the Cold War and the public-private partnership models seen in recent port and high-speed rail infrastructure projects. However, the core assets have shifted from "physical transportation and weapon production lines" to "cloud and AI computing." Data center operators are already in a key position within the global AI ecosystem, and now, through military projects, they further gain a combination of "military computing + long-term contracts + high security levels," elevating their role from ordinary hosting service providers to strategic nodes involved in national security within the AI infrastructure supply chain.
Structurally, this layout represents a triple overlap of "capital concentration + industrial chain reconstruction + regulatory changes": capital is concentrated in a few super-large computing parks on military bases, reshaping the geography and energy landscape of AI training and inference; cooperation between AI companies and the military shifts from cloud service contracts to sharing physical parks and power facilities, changing who controls the underlying computing power and infrastructure pricing; at the same time, Congress is beginning to propose new review and restriction clauses regarding the sources of data center equipment, energy consumption, water resources, and residential safety, transforming AI infrastructure from a general commercial project into a field requiring dual military and environmental oversight, pushing the entire computing power supply chain towards "sovereign computing + security review + long-term capital" restructuring.
ABAB News · Cognitive Law
When computing power bases move into military camps, data centers are no longer just cloud businesses but sovereign infrastructure.
The AI war machine needs not just one or two GPUs, but to turn entire land and energy into long-term contracts.
The moment the military leases the base to private equity, national security issues become tied to fund returns.