Subterranean defense technology company Traysar has teamed up with Edge computing specialist Armada to bring sovereign AI compute underground.

Traysar has launched Terminus, a truck-mounted boring platform that enables infrastructure to be deployed underground.

Traysar and Armada will combine Traysar's expertise in underground systems with Armada's modular data centers. The two plan to explore integrated underground solutions supporting secure AI processing, command and control, communications, and other mission-critical workloads for the US.

The companies said distributed underground deployments can be scaled into a Sovereign AI Grid, meaning workloads never depend on a single site.

“The next generation of critical infrastructure cannot remain exposed. Terminus transforms the Earth’s crust into a protective layer for the systems America depends on most. By combining Traysar’s ability to create and operate in underground environments with Armada’s ability to deploy powerful AI infrastructure virtually anywhere, we are building a new standard for resilient compute—concealed, hardened, and ready to operate when surface infrastructure cannot," said Yadin Soffer, co-founder and CEO of Traysar.

Armada is known for its modular Edge data centers, which can be placed wherever power is available, and the offerings have previously been used by US military departments, including by the US Navy during UNITAS.

Armada's modules come in a variety of sizes, with the company recently launching a 10MW model dubbed Orion, its largest to date. The second-biggest modular offering is Leviathan, with 1.77MW of compute capacity.

“The future of AI infrastructure is distributed, sovereign, and resilient. The Sovereign AI Grid is built so that no critical workload depends on a single site. Terminus adds a new dimension. Together, we can put a fully sovereign, air-gapped AI node beneath the surface, close to the mission it supports, and keep it running when everything above ground cannot," added Dan Wright, co-founder and CEO of Armada.