Armada Raises $230M to Build Deployable AI Data Centers for the Edge

The story of AI infrastructure in 2026 has been largely a story about scale: who can build the largest data centers, secure the most power contracts, and stock the most GPUs. Armada is building a different chapter — one about compute that goes where the hyperscalers cannot or will not.

On May 19, the modular data center company announced a $230 million oversubscribed Series B at a $2 billion pre-money valuation. Co-leads were Overmatch, 8090 Industries, and BlackRock, joined by new strategic investors NightDragon, Johnson Controls, Mitsui, and Singtel Innov8. The round was oversubscribed, reflecting investor conviction in a market that most of the AI infrastructure conversation has overlooked: the edge.

What Armada Actually Builds

Armada’s product is, in concept, simple: a data center in a container. The company designs and deploys modular, containerized compute units that ship to a site, run on local energy, and deliver AI inference and data processing capabilities without any dependency on hyperscale cloud infrastructure.

In practice, the engineering is anything but simple. An Armada unit needs to operate in extreme environmental conditions — the deck of a naval vessel in the North Atlantic, a remote oil rig in the Gulf of Mexico, a forward operating base in an environment with intermittent power. It needs to be physically ruggedized, thermally managed without the precision environmental controls of a traditional data center, and capable of autonomous operation without the on-site IT staff that enterprise facilities take for granted.

The units are, in Armada’s framing, deployable rather than merely modular. The distinction matters: modular data centers like those deployed by large cloud providers are permanent or semi-permanent infrastructure installations. Armada’s are designed to be moved — set up, run, and redeployed — which creates entirely different requirements for physical durability, power systems, and operational software.

The Customer Base Tells the Story

Armada’s existing customers span two categories that converge on the same constraint: they need serious compute, and they are somewhere the internet reliably isn’t.

Defense and national security customers — including the United States Navy — need AI inference capabilities that are not dependent on cloud connectivity that could be disrupted in a contested environment. The ability to run sophisticated AI models (targeting, logistics optimization, reconnaissance image analysis) at the tactical edge, without relaying data to a remote data center, has become a recognized operational requirement. Armada’s units provide it.

Remote industrial operations — offshore oil rigs, mining sites, energy infrastructure in geographically isolated locations — face a different but structurally similar challenge. These facilities generate enormous volumes of sensor data, equipment telemetry, and operational signals that AI can analyze to prevent failures, optimize production, and reduce maintenance costs. But sending all of that data to the cloud for processing is either impractical (bandwidth constraints), expensive (satellite data transmission costs), or unacceptable from a data sovereignty perspective. Processing it on-site requires compute that traditional industrial edge hardware is not designed to provide.

The convergence of these two customer categories around the same product class is a validation signal: the use case is not exotic. It is a fundamental gap in the AI infrastructure stack that scales with the deployment of AI-enabled systems across any environment that lacks reliable, high-bandwidth connectivity.

Johnson Controls and the Manufacturing Bet

The most strategically significant element of the Series B is not the capital amount — it is the manufacturing partnership. Johnson Controls, one of the world’s largest building technology and infrastructure companies, is not just a financial investor in this round. It has signed a manufacturing agreement to produce Armada’s modular units at a new 400,000-square-foot Arizona facility called Galleon Forge One.

This matters because it resolves the bottleneck that has limited Armada’s ability to convert demand into revenue: manufacturing capacity. Until Galleon Forge One, Armada was effectively a boutique product — each unit required significant custom engineering and was produced in limited quantities. The new facility transforms the unit into something an enterprise IT procurement team or a government acquisition program can actually buy at scale, with lead times and unit economics that support large contract structures.

Johnson Controls’ involvement also brings two things Armada would take years to develop independently: credibility with the large industrial and government customers that are Armada’s target market (Johnson Controls already has relationships across defense, healthcare, energy, and government), and deep expertise in building physical infrastructure systems that meet military and industrial certification requirements.

Growth Metrics That Justify the Valuation

Armada’s growth numbers are striking enough to warrant scrutiny. Customer bookings reportedly grew 540% between fiscal year 2025 and fiscal year 2026. More dramatically, Q1 FY2027 alone posted approximately 2,000% year-over-year growth as demand from defense, energy, and industrial customers surged.

Numbers at this scale should always be contextualized. A company growing from a very small base can post large percentage gains without yet reaching the scale that makes those percentages meaningful in absolute dollar terms. The $2 billion pre-money valuation at Series B — a figure that represents real capital commitment by sophisticated institutional investors including BlackRock — is a more grounded indicator that the underlying business is substantive, not just rapidly growing from a trivial base.

The oversubscribed nature of the round, and the specific combination of investors, also signals something about the market structure: BlackRock’s participation reflects infrastructure fund interest (these are long-duration, asset-backed assets), while NightDragon’s involvement (a firm focused on cybersecurity and national security technology) and Mitsui’s participation (the Japanese trading house with major energy and industrial assets) reflect the specific customer verticals Armada is penetrating.

The Broader AI Infrastructure Landscape

Armada’s raise arrives in a context where AI infrastructure investment has become the dominant theme in technology capital markets. NVIDIA’s Vera Rubin chips are scheduled for Q3 2026 delivery at 5x Blackwell performance. Amazon’s Trainium chip business has accumulated a $225 billion backlog. The major cloud providers are in a capital expenditure race that has no near-term ceiling.

All of that investment is, fundamentally, in centralized infrastructure: the assumption that AI workloads will flow to large, connected facilities optimized for training and inference at scale.

Armada is betting on the other half of the market — the workloads that cannot flow to centralized infrastructure because of connectivity constraints, latency requirements, data sovereignty rules, or physical operating environments. It is a market that most AI infrastructure investors have not focused on because it does not fit the hyperscale model, and because it requires deep domain expertise in industries (defense, energy, maritime, mining) that are culturally distant from the technology-first orientation of Silicon Valley.

The Galleon Forge One factory represents a bet that this market is large enough, and the timing right enough, to justify building manufacturing capacity specifically for it. At 400,000 square feet in Arizona, with Johnson Controls as production partner, it is a tangible commitment to the thesis that deployable AI compute is a product category — not an engineering curiosity — and that the demand is real enough to justify industrial-scale production.

What to Watch

The critical question for Armada over the next 12 to 18 months is whether Galleon Forge One can convert the demand signal from defense and industrial bookings into delivered units at a pace that matches the growth projections implicit in the Series B valuation. Manufacturing scale-up is notoriously difficult for hardware companies, and the combination of novel product design with ambitious production timelines is where many hardware startups find their trajectory constrained by execution rather than demand.

The defense market also carries its own rhythms: large government contracts move through acquisition cycles that are not always synchronized with a startup’s capital needs. Armada’s relationship with the U.S. Navy and other defense customers will need to evolve from early deployment partner to program-of-record contractor status if it is to generate the sustained, predictable revenue that the industrial-scale manufacturing bet requires.

If it gets there, Armada will have built the infrastructure layer for the AI workloads that the hyperscalers never could reach — and the $2 billion valuation today will look modest in retrospect.