The Pentagon Is Building a Private-Capital Compute Grid

The most important U.S. defense-industrial story of 2026 may not be a new missile, aircraft or AI model. It may be the financial architecture forming underneath all of them.

Across a series of separate official actions, the U.S. Army and Department of the Air Force are opening military land, energy systems, industrial facilities and test ranges to private capital and commercial operators. No single government release calls this a unified network. SIAIntel therefore uses Military Industrial Compute Grid as an analytical label — not an official program name — for the emerging model in which secure land, electricity, computing capacity, mineral processing, manufacturing and test infrastructure are being connected through public-private structures.

The source trail is unusually explicit. The Army selected Carlyle for a hyperscale data-center project on about 1,384 acres at Fort Bliss and CyrusOne, owned by funds managed by KKR and BlackRock, for approximately 1,201 acres at Dugway Proving Ground. The companies are expected to finance, build, operate, maintain and eventually decommission the facilities on Army land. U.S. Army — hyperscale data-center selections

The Army's own Strategic Capital Initiative says it is shifting away from relying only on traditional government contracting and explicitly seeks private investment across energy resilience, the organic industrial base, logistics, advanced manufacturing, real assets such as data centers and industrial parks, and critical minerals. U.S. Army — Strategic Capital Initiative

The same model is now visible in power. The Army invited private developers to finance, build and operate commercial generation on Army land, with long-term leases and an islanded-resilience requirement for grid outages. U.S. Army — energy resilience tranche

The Air Force has gone even further in defining AI infrastructure scale. Its AI data-center lease framework says qualifying projects include facilities requiring more than 100 MW of new load and at least $500 million of capital expenditure; if onsite generation is included, the Department of the Air Force requires a first right of refusal to that generation. Air Force Civil Engineer Center — AI data-center and energy opportunities

This is not confined to compute and electricity. The Army has conditionally awarded long-term leases for privately financed critical-mineral processing at Anniston Army Depot, Pine Bluff Arsenal, Red River Army Depot and Tooele Army Depot. U.S. Army — critical-mineral processing leases

And the industrial layer is opening as well. Tobyhanna Army Depot signed a first-of-its-kind cooperative arrangement with Darkhive that allows the company to work inside depot facilities and use established production capacity for small-unmanned-aircraft-system development and production. U.S. Army — Tobyhanna/Darkhive production arrangement

30-second intelligence brief

The signal is not that the Pentagon is privatizing military bases. It is that the military is increasingly treating land, power, factories, test ranges and compute access as strategic assets that can attract external capital.

That distinction matters. Traditional procurement asks government to buy a finished product. The emerging model can instead let private operators finance long-lived infrastructure, earn commercial returns and provide the military with rent, resilience, capacity, infrastructure improvements or other in-kind value.

For investors, utilities and technology companies, this changes the boundary of the defense industrial base. The relevant map is no longer just primes, suppliers and appropriations. It increasingly includes infrastructure funds, data-center developers, utilities, energy developers, mineral processors, private credit, industrial software and operators able to work inside secure federal environments.

The first layer: military land becomes an AI infrastructure asset

Fort Bliss and Dugway are the clearest evidence. The Army is not merely buying cloud capacity through a services contract. It is using non-excess military land as an input into commercially driven hyperscale infrastructure.

The Fort Bliss project covers about 1,384 acres. Dugway covers approximately 1,201 acres. The Army said initial operating capability is projected for fiscal 2027 at Fort Bliss and fiscal 2029 at Dugway. The deals were conditional selections when announced, so they should not be described as completed data centers or final leases.

The deeper signal is the capital stack. Carlyle sits at the intersection of infrastructure and private markets. CyrusOne is backed by KKR and BlackRock-managed funds. In other words, an Army modernization requirement is being translated into a form that global infrastructure capital already understands: long-duration land access, large physical assets, energy demand, contracted infrastructure and commercial operation.

That can accelerate buildout without requiring the Army to carry the entire upfront capital burden. It also exposes the model to private-sector discipline: power availability, interconnection, financing cost, construction risk and utilization must still work economically.

The second layer: power is moving onto the same strategic balance sheet

Compute without reliable electricity is not strategic compute. The Army's June energy-resilience tranche makes that explicit.

Private partners are expected to finance the full project lifecycle. The Army provides land, typically through a long-term lease. During normal conditions the generation can sell power to a third-party offtaker. During a sufficiently long grid outage, the installation must be capable of receiving continuous islanded power.

The Army is open to solar, natural gas, geothermal and nuclear; nuclear proposals are capped at 500 MW per unit in that solicitation. This does not mean a 500 MW reactor is being built for a specific data center. It means the Army has created a commercial framework in which generation at utility scale can coexist with military resilience requirements.

On August 5, the Army and Tennessee Valley Authority added another piece. Their new memorandum of understanding focuses on potential energy resources at Redstone Arsenal and Fort Campbell, grid security and rising demand from data centers and manufacturing. U.S. Army — Army/TVA grid-resilience MOU

That is the bridge between the first and second layers: military land is becoming useful not only because it is secure, but because the value of that land increasingly depends on how quickly generation, transmission and resilient power can be attached to it.

The Air Force adds a crucial clause: power optionality

The Air Force AI data-center solicitation makes the power relationship unusually concrete. Qualifying projects must exceed 100 MW of new load and involve at least $500 million of capital expenditure. If a developer includes onsite generation, the Department of the Air Force requires first right of refusal to that generation.

That clause deserves more attention than the headline acreage numbers. It means a commercial AI campus can potentially become a resilience asset for the host installation. The same turbine, storage system, reactor or other generation resource that supports a commercial load can create strategic power optionality for the military.

This is not free power and it is not an automatic transfer. The Air Force says any power obtained would be negotiated separately. But the architecture is important: the military is positioning itself to capture strategic value from infrastructure built primarily with private capital.

The Air Force's Defense Energy Consortium points in the same direction. The consortium model is designed around private financing for energy-resilience projects at installations worldwide, with no guarantee of appropriated funding for future prototype projects. Its published framework explicitly contemplates engagement with investment authorities in host countries including Germany, Japan, South Korea, Turkey and the United Kingdom.

The third layer: minerals move inside the installation perimeter

AI infrastructure and modern weapons both depend on material supply chains. The Army's June 25 critical-mineral lease awards show that military real estate is being used to attack that bottleneck as well.

Four companies received conditional awards to design, finance, build and operate processing facilities on Army installations. The selected sites include Anniston Army Depot, Pine Bluff Arsenal, Red River Army Depot and Tooele Army Depot.

This is structurally similar to the data-center model: the Army contributes access to strategic real estate and mission alignment; industry brings capital, processing capability and commercial execution.

The implication is larger than domestic mining policy. If secure installations become sites where compute, energy, materials and industrial production can coexist, the military base starts to resemble a dual-use infrastructure platform rather than a location dedicated only to military consumption.

The fourth layer: the depot becomes a commercial production interface

Tobyhanna's agreement with Darkhive shows what happens when the same logic reaches manufacturing.

The arrangement allows Darkhive personnel to work inside Tobyhanna facilities and use depot equipment and production capacity for small-drone development and production. The Army describes the arrangement as a first use of the relevant statutory authority by Tobyhanna and says it can generate working-capital funds for the depot.

This is not the outsourcing of the depot. The more precise reading is that the boundary between the organic industrial base and commercial defense technology is becoming more permeable.

That matters for drones because iteration speed is a strategic variable. A private company that can design, test, modify and produce using an established military industrial facility can potentially shorten the distance between field feedback and scaled output.

The fifth layer: military test ranges become part of the innovation stack

On August 7 the Army opened four major U.S. test ranges and collaborated to open an international range in Morocco to private industry for internal R&D and rapid experimentation. U.S. Army — private-industry test-range access

The designated locations include Dugway Proving Ground, West Cibola Range at Yuma Proving Ground, Camp Shelby and Camp Grayling, plus the Morocco multidomain range complex. Companies do not need an existing federal contract merely to request access.

This closes an important loop. A commercially backed firm can increasingly see a pathway from private financing, to secure military test environments, to military-linked production capacity, while adjacent infrastructure projects expand power and compute.

That does not guarantee procurement. It does reduce some of the institutional friction that traditionally separated commercial technology cycles from military infrastructure.

Why SIAIntel calls this a Military Industrial Compute Grid

The phrase is deliberately analytical. There is no official Pentagon program carrying that name.

But the underlying components now line up:

Land provides secure, large-scale sites.

Power provides resilience and the energy density required by AI and advanced manufacturing.

Compute provides training, inference, simulation and digital engineering capacity.

Materials provide the physical inputs for electronics, motors, magnets, batteries and weapons.

Factories turn designs into deployable hardware.

Ranges provide rapid testing in realistic and contested environments.

Private capital links the pieces by financing assets that can have both commercial and national-security value.

The word “grid” therefore refers to an economic and industrial network, not a claim that the Pentagon has created one centralized electrical or computing grid.

The hidden financial shift: from procurement budget to infrastructure economics

This model changes who must understand defense demand.

In a classic acquisition cycle, a defense contractor watches appropriations, program offices and contract awards. In a private-capital infrastructure cycle, a much wider group must model military demand: infrastructure funds, lenders, utilities, developers, equipment suppliers and real-estate operators.

Their questions are different. Is the lease long enough to amortize construction? Who owns interconnection risk? Can commercial offtake support debt service? What happens if the military mission changes? What security restrictions affect utilization? Who pays for transmission upgrades? What form of in-kind consideration has measurable value?

This is where SIAIntel's earlier Dollar-Watt framework intersects the defense story. Capital cost and power cost do not disappear because a project sits on federal land. They become part of the national-security infrastructure equation.

The most important asset may be optionality, not ownership

The military does not need to own every data hall, generator or production machine to derive strategic value from it.

Long leases, first-refusal rights, islanding requirements, in-kind consideration, secure access and shared production authority can create optionality. In a normal market environment, assets can serve commercial demand. In a disruption, some can provide capacity, power, production or testing access that improves military resilience.

That is economically attractive because idle strategic redundancy is expensive. Dual-use infrastructure can potentially keep assets productive in peacetime while preserving contractual pathways for national-security use.

The risk is that optionality only has value if contracts, interconnection rules, cyber security, physical security and operational priorities work under stress. A right written on paper is not the same as deliverable capacity during an emergency.

What this means for utilities and grid operators

Military installations may become larger nodes in regional power planning.

A >100 MW AI data center is already a large-load problem. Add onsite generation, manufacturing and mission-resilience requirements, and the interconnection becomes more complex. The Army/TVA MOU explicitly cites growing demand from data centers and manufacturing as part of the reason to examine new resources at Redstone and Fort Campbell.

Utilities therefore face a dual mandate: serve new commercial load while preserving military reliability. Projects that can bring their own generation or finance network upgrades may become more attractive than projects that simply request load service.

This is a potential competitive advantage for installations with strong transmission access, available land, predictable permitting pathways and nearby industrial ecosystems.

What this means for AI infrastructure investors

The military-land model creates a new asset class candidate, but not a risk-free one.

The upside is obvious: large secure parcels, federal mission relevance, potential power-development options and long-lived infrastructure demand. The downside is equally real: security controls, environmental review, uncertain final lease terms, grid queues, political change and restrictions on who can finance or operate sensitive assets.

Investors should therefore avoid treating “on military land” as a substitute for underwriting. The strongest projects will still require credible tenants, executable interconnection, bankable lease terms and realistic construction timelines.

The Fort Bliss and Dugway announcements are conditional negotiations, not completed lease closings. That boundary is critical.

What could invalidate the thesis

The Military Industrial Compute Grid thesis would weaken if these initiatives remain isolated pilots rather than repeatable models.

It would also weaken if private capital proves unwilling to accept the security, permitting and mission constraints; if utilities cannot provide power at scale; if Congressional or legal constraints narrow lease authorities; or if commercial utilization cannot support project economics.

A second failure mode is fragmentation. Data centers, power, minerals, depots and ranges may all expand without ever becoming economically interoperable. In that case the correct description would be a collection of modernization projects, not a coherent infrastructure network.

SIAIntel therefore treats the thesis as OPEN, not confirmed.

The next confirmation signals

The strongest evidence will not be another speech. It will be contracts and operating structures.

Watch for final lease terms at Fort Bliss and Dugway; named power suppliers and interconnection agreements; commercial offtakers for Army generation projects; additional installations using the same lease model; explicit compute or infrastructure in-kind consideration; new depot-industry agreements after Tobyhanna; and repeated private use of Army test ranges.

Also watch the Office of Strategic Capital. OSC's formal mission is to attract and scale private capital into national-security technology and supply chains, making it another potential bridge between defense priorities and private balance sheets. Office of Strategic Capital

SIAIntel verdict

CONFIRMED FACT: The Army and Air Force are separately using private-finance, lease and partnership structures for data centers, energy resilience, critical minerals, industrial production and testing.

SIAINTEL ANALYTICAL THESIS — OPEN: These separate structures are beginning to form a broader Military Industrial Compute Grid in which military installations become platforms for dual-use compute, power, materials and manufacturing.

WHAT WOULD CONFIRM IT: Repetition across installations, final commercial terms, shared infrastructure economics and explicit links between compute, power and industrial capacity.

WHAT WOULD REJECT IT: Projects remain isolated, financing fails to close, grid constraints dominate, or the military retains no meaningful access or resilience value from privately financed assets.

The strategic shift is not that private capital is replacing the defense budget. It is that the U.S. military is learning to use secure land, demand, legal authorities and mission relevance as capital-forming assets. If that model scales, the next defense-industrial race will be fought not only over who builds the best weapons, but over who can finance and connect the compute, power, materials and factories underneath them.