Carlyle and KKR Build $4B U.S. Army Data Centers

Carlyle and KKR are reported to be partnering to build two data centres for the U.S. Army at roughly $2 billion apiece, a Financial Times scoop published on March 26, 2026 and republished by Seeking Alpha indicates. The two facilities would represent approximately $4 billion of capital expenditure tied directly to a U.S. military customer, an uncommon scale for private-sector-led builds for a single federal client (Financial Times, Mar 26, 2026). The transaction — if finalized under the reported structure — would sit at the intersection of private equity infrastructure strategies, sovereign IT modernisation, and the growing trend of defence agencies outsourcing large-scale cloud and edge compute projects. This note sets out the context, data-driven implications for the data-centre and private-equity markets, key risks, and a contrarian Fazen Capital perspective on what the transaction would signify for long-duration infrastructure allocations.

Context

The reported FT story (Mar 26, 2026) places two $2 billion builds for the U.S. Army in the context of accelerating demand for secure, sovereign compute capacity. Over the past five years the Department of Defense has increasingly prioritized resilience, onshore capacity and secure cloud-like environments following the cancellation of earlier single-vendor initiatives and a shift toward multi-supplier models. Governments and defence agencies now often require facilities that combine physical security, classified enclaves and robust connectivity to both public cloud providers and tactical networks.

Private capital has been active in data-centre infrastructure since the mid-2010s, but involvement at the sovereign-defence level has been sporadic. The reported scale — $4 billion total across two assets — is materially larger than typical commercial colocation projects, which commonly range from $50 million to $500 million depending on site, capacity and land costs. The FT report suggests a potentially bespoke procurement vehicle that could blend equity, project finance and long-term service agreements, a structure that fits recent private-equity infrastructure plays aiming for annuity-like cashflows and government-contracted revenues (Financial Times, Mar 26, 2026).

This activity also follows a broader macro backdrop of elevated capex costs and supply-chain pressures. Construction input inflation since 2021 has raised data-centre build costs globally; industry benchmarks for brownfield-to-hyperscale greenfield buildouts commonly sit in the $8–$12 million per MW range depending on land and connectivity — a relevant comparator when assessing the implied scale of a $2 billion facility.

Data Deep Dive

The published figures are clear: two sites at $2 billion each equals $4 billion of reported project value (Financial Times; Seeking Alpha, Mar 26, 2026). Translating dollar value into capacity is not a linear exercise, but using a conservative industry benchmark of $8–$10 million per MW for full turnkey builds suggests a single $2 billion centre could imply 200–250 MW of gross IT load when accounting for site works, redundancy, security overlays and fibre connectivity. That rough conversion provides a sense of scale: combined, the two centres could represent roughly 400–500 MW of secure capacity if built to hyperscale-class specifications.

By comparison, many commercial edge facilities are measured in single-digit MW increments; the reported Army facilities would therefore be multiple times larger than a typical military regional edge node and sit more closely to campus or hyperscale footprints. Publicly traded data-centre REITs (for example, large hyperscalers) have historically sought 50–200 MW campus builds to support multi-tenant and single-tenant strategies. The reported centres would be among the more capital-intensive, single-client projects undertaken by private owners in recent years.

Timing on the deals is not public; FT did not disclose firm execution dates or contractual form. For investors and counterparties, the execution risk includes permitting timelines, grid interconnection queues, and long-lead items (generators, chillers, specialised security hardware). Those elements have driven delivery schedules of 24–48 months for large brownfield expansions historically; greenfield sites with heavy security requirements and multiple carrier connections can extend beyond that window.

Sector Implications

If Carlyle and KKR complete these builds, it would reinforce private-equity appetite for brownfield-plus-greenfield infrastructure tied to public-sector contracts. The implication for the data-centre sector is twofold: first, private capital would likely push more bespoke facilities into the market that meet sovereign security standards, and second, public-data-centre operators could face differentiated competition where military-grade requirements limit standard colocation offerings. Public REITs and operators such as Equinix and Digital Realty may be less active in classified or near-classified defence work due to compliance burdens and potential reputational risk, opening a niche for specialised private owners.

For private-equity investors, these projects offer potential annuity-like cashflows if structured as long-term availability payments, but they also compress traditional yield arbitrage: higher security and compliance premiums offset some investor return expectations. The deal would also set a comparability point for future agency procurements; a $4 billion build for one branch can serve as a de facto benchmark in discussions across other DoD services and federal agencies considering similar onshore compute expansion.

On a broader market level, the transaction highlights the shift of defense IT spend from operational procurement to capital deployments that are persistent and infrastructure-like. That shift could reframe defence IT as an investible asset class, blurring lines between defence contracting and infrastructure investing. Market participants will watch contract terms (capex recovery, service-level agreements, indemnities) for precedent.

Risk Assessment

Construction and delivery risk is primary. Large-scale, secure data-centre builds involve complex permitting, high-voltage interconnection agreements, and extensive physical security work. Given recent supply-chain disruptions, long-lead electrical and mechanical components could trigger delays and cost overruns. Inflation on labour and materials remains an upside risk to final project cost relative to initial $2 billion estimates if bids are fixed but subcontractor costs escalate.

Contractual risk with a sovereign client is asymmetric. While government counterparties reduce commercial counterparty risk, they introduce political and procurement risk: changes in policy, shifts in defence priorities, or oversight that requires renegotiation can affect revenue profiles. Long-term availability-based contracts mitigate occupancy risk but can embed penalty regimes for failing to meet classified security standards or operational uptime required by the Army.

Finally, cybersecurity and regulatory risk are significant. Facilities intended for government or classified workloads will be subject to cyber accreditation regimes, continuous monitoring, and potentially restrictions on third-party vendor access. Those constraints limit aftermarket monetisation (no general commercial leasing in secure zones) and increase operational complexity and cost.

Fazen Capital Perspective

Fazen Capital's view is contrarian relative to headline narratives that treat this report solely as a win for private capital. We observe that while headline capex is large, true investor returns will hinge on contract structure more than initial build cost. If the Army contract is availability-based with modest margin uplift and heavy indemnities, private owners may accept lower near-term yields for the strategic benefit of establishing a sovereign-sector track record. Alternatively, a structure that permits commercial backfill or hybrid tenancy outside classified zones would materially improve return profiles but may be difficult given security constraints.

We also note a second-order effect: firms that can integrate secure cloud services, carrier neutrality and a modular edge architecture stand to capture recurring revenue beyond base facility hosting. Thus, the strategic value may reside less in the raw asset and more in the operating platform and long-term service agreements. That implies PE sponsors who combine capital with a differentiated operating partner (or carve-out specialised platforms post-build) will likely outperform those treating the project as a pure industrial play.

For institutional investors evaluating allocations to data-centre infrastructure, this transaction underscores the need to parse contractual economics and operational optionality. A $4 billion headline can mask low-margin, high-compliance revenue streams; due diligence must focus on service terms, revenue indexing, and decommissioning or repurposing clauses.

Outlook

Market participants should expect heightened activity in government IT infrastructure procurement over the next 12–36 months. If the reported Carlyle-KKR initiative proceeds, other private managers will likely pursue analogous opportunities as agencies seek onshore, resilient compute capacity. Key near-term catalysts to watch are formal contract awards, joint-venture vehicle announcements, financing syndication, and state-level approvals where the centres will be sited.

Valuation implications for public data-centre companies are likely modest in the near term: most public operators lack the product mix for classified, sovereign builds and therefore will not be direct comparables. However, successful private-sector deliveries could create a secondary market for specialised sovereign-grade assets, with a valuation premium for proven security and long-duration government contracts.

Finally, procurement transparency will shape secondary-market interest. If contract terms allow for predictable cashflows (escalators linked to CPI, multi-decade availability payments), the assets could trade at infrastructure-like multiples; if revenues remain tightly beholden to variable service demand or restricted tenancy, investors should expect discounted valuations to account for operational inflexibility.

FAQ

Q: How might these builds affect public cloud providers' relationships with the DoD?

A: The reported builds are complementary rather than substitutive in many scenarios. Public cloud providers often require sovereign onramps and physically secure interconnection points; private-sector-built facilities can serve as neutral points of presence that federate connectivity to multiple clouds while meeting DoD security needs. This dynamic can increase demand for carrier-neutral interconnection and hybrid-cloud integration services.

Q: Could these centres be repurposed for commercial use if DoD demand falls?

A: Repurposing is technically feasible but commercially constrained. Classified zones and security perimeters can be isolated and decommissioned, but doing so reduces remaining usable floorplates and can require material remediation. Commercial repurposing improves exit optionality for investors but may require upfront contractual permissions and regulatory approvals.

Bottom Line

The FT report that Carlyle and KKR are developing two $2 billion data centres for the U.S. Army signals a notable shift of private capital into sovereign-grade compute infrastructure; execution will determine whether the headline $4 billion translates into durable, infrastructure-like returns or a high-cost specialised asset with limited monetisation optionality. Disclaimer: This article is for informational purposes only and does not constitute investment advice.