Armada Unveils Factory-Built Superintelligence Infrastructure on NVIDIA DSX POD, Targets 500MW With Johnson Controls in 2027

Armada, a specialist in sovereign superintelligence infrastructure, used NVIDIA GTC in Berlin to announce support for the NVIDIA DSX POD reference design, a move aimed at shifting the repetitive construction of AI factories off the job site and into the manufacturing plant. The company said it will engineer, manufacture, integrate and deploy standardized, factory-integrated containers built on the NVIDIA Vera Rubin platform, with cabling, wiring, piping and functional validation completed before any unit leaves the factory.

The announcement matters because the infrastructure underpinning the race toward superintelligence — AI systems that match or exceed human experts across most fields — remains bound to one of the slowest-moving models in the economy. A conventional 50MW facility can take years to become IT-ready and consume roughly 600,000 on-site labor hours, at a time when the construction industry faces a projected shortfall of 288,000 workers by 2027. Permits, power interconnects and community approvals stretch timelines further, leaving companies and nations that already own the chips, land, power and capital waiting for a building.

Under the collaboration, Armada and Johnson Controls plan to build 500MW of distributed sovereign superintelligence capacity in 2027. The partnership combines Armada's experience deploying compute in extreme environments — including frozen Alaska, a U.S. Navy littoral combat ship, desert sites in the Middle East and, soon, a North Sea oil platform — with Johnson Controls' thermal management technology, engineering capabilities and global manufacturing scale. Armada, founded in 2022 and headquartered in San Francisco, produces its modular data centers at Galleon Forge One in Gilbert, Arizona, and Galleon Forge Two in Dublin, Ireland.

Armada's involvement with NVIDIA dates to its acceptance into the NVIDIA Inception program in 2024, when it began working with the company's network simulation technologies. It now draws on the NVIDIA DSX platform across the full life of a superintelligence factory: DSX Air to simulate the facility and validate its Bridge GPU management software before hardware ships, the DSX POD reference design to manufacture containers, and DSX MaxLPS to maximize token output within each site's power envelope once operational. The Armada Platform also supports NVIDIA AI Grid, allowing telecommunications operators, service providers and enterprises to deploy, operate and monetize geographically distributed AI infrastructure for latency-sensitive, real-time workloads.

"The SI race isn't going to be won by whoever pours concrete the fastest," said Dan Wright, co-founder and CEO of Armada. "Companies and nations can have the chips, the land, the power, and the ambition, and still be told to wait years for a building. The NVIDIA DSX POD reference design lets us cut that wait out of the equation."

Charlie Boyle, vice president of DGX Systems at NVIDIA, said demand for compute is outpacing the speed at which the world can build facilities to house it, adding that Armada is positioned to bring Vera Rubin-based factories to customers "who need capacity now, wherever their energy and data are." Joakim Weidemanis, CEO of Johnson Controls, said the combination of the two companies' expertise delivers high-performance infrastructure engineered to bring capacity online quickly where and when it is needed.

Armada deploys sovereign capacity as a grid rather than a monolith: a network of modular factories sited close to energy sources, including stranded, renewable and off-hours power, with minimal community footprint and water impact. Each site is built on an open architecture with zero cloud dependencies, so customers own and control their models, data and the infrastructure that runs them. For the industry, the model signals a broader shift toward prefabricated, standardized AI capacity that can be scaled by adding containers as demand grows — potentially compressing deployment timelines from years to months and reshaping how nations and enterprises pursue sovereign compute.