NVIDIA Announces DSX POD Modular Vera Rubin AI Factory Reference Design
Редакция ЦОД:Главное
NVIDIA has unveiled DSX POD, a reference design for modular, containerized AI factories based on NVIDIA Vera Rubin systems. The company says partners will pre-integrate and validate the modules before shipment, leaving site teams to connect them to prepared power, cooling, network and supporting infrastructure.
- DSX POD is an announced reference design for containerized AI factories based on NVIDIA Vera Rubin systems.
- Modules are intended to be integrated and functionally validated before shipment.
- NVIDIA claims deployment at a prepared site can be reduced from six months to as little as 14 days.
- Prepared power, cooling, network and supporting infrastructure remain prerequisites for on-site deployment.
- NVIDIA’s partner ecosystem is targeting worldwide deliveries in early 2027.
Announced at GTC Berlin, NVIDIA DSX POD is a reference design for complete modular AI factories using NVIDIA Vera Rubin systems. Partners are expected to assemble, connect and functionally validate the containerized modules before they are delivered to a customer site.
Under NVIDIA’s proposed deployment model, the remaining on-site work is the connection of a completed module to pre-established power, cooling, network and other supporting infrastructure. NVIDIA says this approach can reduce deployment time at a prepared site from six months to as little as 14 days.
The design moves integration of compute, networking, cabling, rack-level power and liquid cooling into a factory environment. For operators building AI capacity, this could reduce dependence on scarce on-site integration labor and make delivery of repeatable infrastructure modules more scalable. It does not, however, remove the requirement to secure and build the underlying site infrastructure before delivery.
NVIDIA identified Bull, Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv as part of the global ecosystem working on the offering. The partner ecosystem intends to deliver modular AI factories worldwide in early 2027.
The announcement is an indication of a broader shift toward factory-integrated AI data-center infrastructure rather than extended sequential assembly in the field. Whether the claimed deployment timeline is achievable in commercial use will depend on the readiness of power, cooling, networking and other supporting systems at each site.
DSX POD shifts a substantial share of compute, networking, cabling, rack power and liquid-cooling integration from the construction site to the factory. If the model performs as claimed in commercial deployments, it could shorten AI infrastructure delivery cycles and intensify competition among modular infrastructure, power and liquid-cooling suppliers—while leaving grid, cooling and other site-enablement work as critical prerequisites.
- NVIDIA — Newsroom(первоисточник) / АнглийскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.