NVIDIA Unveils Containerised DSX Pod AI Factory, With Shipments Planned for Early 2027
Редакция ЦОД:Главное
NVIDIA has introduced the DSX Pod AI Factory, a containerised system integrating seven Vera Rubin racks with factory-installed cabling and liquid cooling. The company says the approach can reduce AI data-centre installation time from six months to 14 days; shipments are planned for early 2027.
- DSX Pod integrates seven Vera Rubin racks in a transportable containerised module.
- NVIDIA says factory integration of cabling and liquid cooling can cut installation time to 14 days from six months.
- The model could reduce on-site integration work, provided power and network infrastructure are ready at delivery.
- NVIDIA plans production with Foxconn, Gigabyte, Lenovo, HPE, Quanta and Pegatron.
- Shipments are planned for early 2027; pricing has not been disclosed.
NVIDIA unveiled its DSX Pod AI Factory at GTC Berlin 2026 on 21 October. The containerised AI infrastructure product combines seven racks of the Vera Rubin system in a single transportable module.
NVIDIA plans to complete the installation of equipment, cabling and liquid-cooling infrastructure at the factory before the module is transported to the deployment site by conventional road transport. The company states that this pre-integration can reduce the installation period for an AI data centre from six months to 14 days.
For data-centre operators, the model could shift a larger share of integration activity from the site to the manufacturing process. This may reduce the amount of on-site systems integration required, although it also places greater importance on having the receiving site’s power supply, grid interconnection and network connectivity ready before delivery.
NVIDIA plans to manufacture DSX Pod with Foxconn, Gigabyte, Lenovo, HPE, Quanta and Pegatron. The system weighs about 28 tonnes. Shipments are planned to begin in early 2027; this is a planned delivery schedule rather than evidence of commercial deployment or operational performance.
NVIDIA has not disclosed DSX Pod pricing. The announcement is based on a single source, and the company’s 14-day installation claim does not specify site conditions, the scope of work included, or the methodology used for comparison with the six-month timeframe.
The product signals a move from discrete GPU delivery towards standardised, pre-integrated AI infrastructure. If NVIDIA’s installation-time claim is achieved in deployment, it could reduce on-site integration work and make site power, grid interconnection and network readiness more critical determinants of time to service.
- — / 22 октября 2026 г. · КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.