The design is built for the Vera Rubin platform and is delivered through NVIDIA ecosystem partners.
The company says partners can complete assembly, rack-to-rack cabling, wiring, piping, and functional validation before shipment. Standardized units are then connected at a prepared site to power, facility water, cooling, and network infrastructure. NVIDIA explicitly says the design does not eliminate onsite work.
DSX POD is paired with NVIDIA DSX MaxLPS, which the company describes as software for policy-driven power control, dynamic workload orchestration, and telemetry within defined power limits. NVIDIA says the initial focus is greenfield, liquid-cooled infrastructure, including NVIDIA Cloud Partners.
The source does not provide a price, a commercial availability date, a fixed hardware configuration, named delivery partners, or independently measured deployment-time or efficiency results. It also does not substantiate the broader set of GTC Berlin claims included in the discovery material.