The Pulse
NVIDIA Says DSX POD Can Cut AI Factory Deployment to 14 Days
NVIDIA introduced DSX POD, a partner-built reference design for containerized AI factories based on its Vera Rubin platform. The company says connecting a factory-integrated unit to a prepared site could cut deployment from six months to as
AI.info Team · Written with AI · fact-checked ·
NVIDIA says its new DSX POD design could shrink the time to deploy an AI factory at a prepared site from six months to as little as 14 days. Announced October 21 at GTC Berlin, the design puts computing, networking, cabling, rack-level power and liquid cooling into modular containers that partners integrate and test before shipment.
The containers are designed around NVIDIA’s Vera Rubin platform and arrive ready to connect to site power, cooling, networking and other supporting infrastructure. The 14-day figure describes deployment at a prepared site, not the time required to secure land, build supporting utilities or complete every stage of a data-center project.
DSX POD moves integration into the factory
Traditional projects bring much of the equipment installation and system integration to the customer site. NVIDIA’s design shifts more of that work to manufacturing facilities, where partners can complete rack-to-rack cabling, wiring, piping and functional checks before the units travel to their destination.
That work can happen while site teams prepare foundations, power and external connections. After delivery, the remaining steps include placing the containers, connecting them to each other and to site systems, checking for transport damage, and commissioning the installation. NVIDIA’s technical description says the model reduces field integration; it does not remove the need for site preparation or commissioning.
Huang ties speed to production revenue
For NVIDIA, the commercial case rests on how soon a new facility can begin running customer workloads. The company argues that an earlier start to model training and production inference gives operators more time to sell computing capacity, though the announcement does not provide independent measurements of deployment times or revenue effects.
“An AI factory creates value when it starts producing tokens — not when construction is complete,”
Jensen Huang, founder and CEO of NVIDIA, in the NVIDIA announcement
DSX POD is a reference design, rather than a requirement for every NVIDIA-based data center. NVIDIA’s technical blog says conventional construction and other onsite-integrated approaches remain options; the containerized design offers a standardized route for operators who prioritize repeatable deployments and less onsite assembly.
Seventeen partners target early 2027
NVIDIA says 17 companies are working with it to deliver DSX POD AI factories worldwide in early 2027. The named partners include Anduril, Armada, Bull, Crusoe, Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv, alongside Delta, Eaton, Foxconn’s Ingrasys unit, GIGABYTE, DayOne, Pegatron, Quanta Cloud Technology and Wistron.
The arrangement leaves manufacturing and deployment with the partner ecosystem. NVIDIA supplies the reference design; companies in the group bring computing systems, modular construction and infrastructure expertise. That structure makes availability dependent not only on the design, but also on partners’ ability to build the units and on each customer site being ready to receive them.
Fourteen days starts after the site is ready
The shorter schedule depends on moving work earlier, not skipping it. The site still needs foundations, utility connections and external network access, while delivered systems require connection and final checks. NVIDIA’s own technical account says partners adapt power and cooling to local requirements and commission the full installation after delivery.
The announcement sets an early-2027 target for partner deliveries but does not name specific customers, deployment locations or unit counts. Those details will show whether the factory-built approach can translate its stated 14-day site deployment into operating AI capacity across different projects.