Armada raises $230M at $2B to bring modular AI data centers to the 70% of the world without hyperscale infrastructure

Armada raises $230M at $2B to bring modular AI data centers to the edge

Armada builds modular AI data centers for the 70% of the world that lacks hyperscale infrastructure. CEO Dan Wright's pitch is simple: AI is useless without compute nearby, and most of the world doesn't have it.

The company raised $230M at a $2B pre-money valuation earlier this year and is now manufacturing its modular units, called Galleons, at a facility called Galleon Forge One in Gilbert, Arizona, built with partner Johnson Controls.

The latency case is concrete. The state of Alaska had 28 hours of latency processing drone data for avalanche and flood emergency response before Armada deployed its first data centers there. The Navy operates with similar constraints in the middle of the ocean. Wright's argument is that data generated far from processing infrastructure becomes worthless for time-sensitive decisions, whether that's a battlefield scenario, a flood, or an avalanche.

Sovereignty is the second driver. Governments and enterprises want to fine-tune the latest models on sensitive data sets without sending them to the cloud. Armada's Galleons enable what Wright calls Sovereign AI factories, where customers can ingest open-source and commercial models, fine-tune them locally on proprietary data, and push updated models to distributed edge nodes. The company also supports federated learning, where model fine-tuning happens at the site level and feeds back into a core model, improving it without centralizing the data.

“Armada is the hyperscaler for the edge. We build the infrastructure for the 70% of the world that doesn't have the big hyperscale data centers today... We just raised earlier this year $230,000,000 at a $2,000,000,000 pre. We launched Galleon Forge One which is a factory with our partner Johnson Controls in Gilbert, Arizona.”

Form factors and power

Armada offers three Galleon form factors. The original sub-megawatt units handle edge inference. Leviathan, announced last year, runs at 2 megawatts per unit. The newest, Orion, announced at the New York Stock Exchange, runs at 10 megawatts per unit. Wright says that combination makes Armada the fastest path from zero to 200 megawatts anywhere in the world.

Power is sourced from stranded energy rather than new grid connections. A Norwegian customer, Fossifal, runs Galleons on 99% hydroelectricity across distributed sites in Norway, Finland, and Sweden, with plans to scale past a gigawatt. In Australia, a partner called WindDC is tapping stranded wind and solar, in a market where 7.2 terawatt hours of energy were curtailed last year because the grid was overloaded while stranded power sat unused.

Wright is direct that hydroelectricity and stranded energy broadly are underappreciated as AI power sources relative to the attention nuclear and solar receive.

Chip flexibility

Because Galleons are manufactured rather than constructed, Armada can iterate on hardware configurations quickly. Customers currently running Nvidia GB-series chips are already evaluating Vera Rubins and chips from other vendors. Wright says Armada's model lets customers experiment with different configurations, order more of what works, and scale with actual demand rather than overbuilding for a workload they haven't validated yet.

Armada is currently at a $2B valuation on $230M raised, with paying customers that include the U.S. Navy, the state of Alaska, and Fossifal in Norway.