Science & Tech · AI · 3 hrs ago
Nvidia unveils modular AI data center design, with deliveries due in 2027
Nvidia has introduced DSX POD, a modular design for building AI data centers.
It combines the company’s upcoming Vera Rubin computing systems with networking, electrical equipment and liquid cooling inside factory-assembled containers.
Nvidia says prepared sites could be set up in 14 days instead of six months, but they must already have the required power, cooling and other infrastructure.
The company also claims its closed-loop cooling could save water and that one technology could deliver the same computing output with 30% less power.
These efficiency estimates depend on facility design and operating conditions, and need independent verification in real-world use.
Nvidia says deliveries with manufacturing partners are expected to begin in early 2027.
Nvidia introduced DSX POD, a modular design that combines its upcoming Vera Rubin computing systems with data center equipment in factory-assembled containers.
The company says deployment could take 14 days instead of six months if a site has the required supporting infrastructure ready.
The design includes computing hardware, networking, electrical distribution and liquid cooling.
Nvidia says its cooling approach can save water, and its DSX MaxLPS technology can provide equivalent token output using 30% less power.
Deliveries are expected to begin in early 2027.
- Who
- Nvidia introduced the DSX POD design, with manufacturing and supply partners including Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv.
- What
- A modular data center design that combines Vera Rubin systems with networking, electrical distribution and liquid cooling in factory-assembled containers.
- When
- Deliveries are expected to begin in early 2027.
- Where
- Not stated.
- Why
- Nvidia says factory assembly could speed deployment and bring AI capacity online sooner. It also claims the design can improve energy and water efficiency.
This story does not have two clearly opposing sides.
TechRadar reported that Nvidia had introduced its DSX POD modular infrastructure design.
Deliveries are expected to begin.
- Deployment estimate
- 14 days, compared with six months, if supporting infrastructure is ready
- Reference system
- Vera Rubin NVL72
- Cabling
- Approximately 17 kilometres of Spectrum-X cabling
- Water savings claim
- 54 million litres daily at a 2GW deployment compared with conventional cooling-tower systems
- Power-efficiency claim
- Equivalent token output using 30% less power with DSX MaxLPS
- Expected deliveries
- Early 2027
What people are saying
An AI factory creates value when it starts producing tokens — not when construction is complete.
With NVIDIA DSX POD, we’re bringing factory-built precision and speed to AI infrastructure construction…so customers can scale capacity and bring revenue-generating production online faster.