DDN’s Enterprise AI HyperPOD is part of AMPHIX, a modular edge AI facility being developed by Strata Expanse on the tribal lands of the Cachil Dehe Band of Wintun Indians in Colusa County, Northern California. The deployment brings together site infrastructure, power, compute, storage and orchestration in a containerized AI Factory-in-a-Box model.
The facility is designed to make an integrated AI environment available for customer proof-of-concept testing. RAVEL, whose AMPHIX AI Infrastructure Platform provides the orchestration layer, says the system has been built and tested. The announcement describes the installation as ready for customer POCs; it does not identify an end-user customer.
Combining power and compute infrastructure
Strata Expanse supplies the physical foundation for AMPHIX, including land, power, cooling, connectivity and supporting infrastructure. The company, a real estate investment trust founded in 2023 and headquartered in The Woodlands, Texas, develops modular infrastructure for AI and high-performance computing. Its model is based on owning, operating and maintaining non-IT infrastructure through long-term subscription agreements described as “Grey Space as a Service.”
At the Northern California site, Strata Expanse is working with Colusa Indian Energy (CIE), a 100 percent tribally owned clean-power infrastructure company, on microgrid and energy infrastructure. Plans are underway to expand on-site generation, incorporating solar, battery storage and generation, to exceed 100 MW within 18 months. The stated goal of the AMPHIX model is to pair this site-level power and infrastructure with compute hardware and software, reducing dependence on conventional utility-grid capacity.
Supermicro supplies the system hardware for AMPHIX, while RAVEL provides its orchestration software. DDN’s contribution is its Enterprise AI HyperPOD, which is based on the company’s Infinia object storage. The equipment is built and supplied by Supermicro. Together, the components form a pre-integrated environment intended to be evaluated as a complete stack rather than as separate infrastructure products.
HyperPOD testing and deployment context
DDN describes HyperPOD as a pre-validated, modular AI Factory based on the NVIDIA Data Platform architecture. It is offered in predefined sizes and is designed for rapid deployment. At AMPHIX, RAVEL says it stress-tested HyperPOD with both inference and training workloads.
RAVEL reported that a 24-hour training test sustained 98.8 percent GPU utilization, with no thermal throttling or GPU errors and 37°C of thermal headroom at peak load. These are results reported by RAVEL; the announcement does not provide independent test validation or further details about the test configuration. RAVEL said the results provide performance evidence for partners and prospective customers assessing infrastructure options.
RAVEL Chief Product Officer Philippa Carrol said the company views HyperPOD as a packaged, production-ready solution created through collaboration among DDN, NVIDIA, Supermicro and RAVEL. She said the AMPHIX environment had been built and tested and would be made available for customer POCs, allowing organizations to test workloads and validate configurations before making production decisions.
The Northern California project follows Strata Expanse’s first AI Factory-in-a-Box, an AI Center of Excellence established in Ironton, Ohio, in December 2025. That earlier facility was part of a planned 30-site national rollout. Supermicro supplied its IT equipment, including liquid- and air-cooled clusters and GPU and CPU platforms for large language models, generative AI, training and HPC. The storage supplier for the Ohio installation was not identified.
Strata Expanse has described a potential network of up to 100 sites. That figure represents the possible scale of its network, rather than a confirmed number of operating AMPHIX facilities. The Colusa County project instead highlights the proposed combination of modular compute, orchestration and locally supplied power, with customer testing intended to precede broader production use.