Nvidia unveils DSX POD modular AI factory promising 14-day deployments and lower resource use

Nvidia announced the DS X POD, a factory-built modular AI “factory” that integrates its upcoming Vera Rubin NVL72 systems, networking, power distribution and liquid cooling inside shipping containers TechRadar Middle East. The company claims the approach can shrink deployment schedules from six months to 14 days, assuming the destination already has power, cooling and networking infrastructure in place.

Each container arrives pre-cabled, with cooling pipes, power connections and Spectrum-X cabling linking seven liquid-cooled racks. Nvidia CEO Jensen Huang linked the speed to earlier revenue generation, noting that “an AI factory creates value when it starts producing tokens — not when construction is complete.”

Beyond speed, Nvidia promotes resource efficiency. For a 2 GW deployment, the closed-loop liquid cooling system is said to avoid water evaporation, saving an estimated 54 million litres of water daily compared with traditional cooling towers. The DS X MaxLPS technology allegedly delivers the same token output while using 30 % less power, and DS X Flex can modulate consumption based on electricity prices and grid conditions.

Deliveries are slated for early 2027, with manufacturing partners Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv.

Our read: The DS X POD’s factory-assembled model could reshape AI infrastructure rollouts by moving integration off-site, but actual energy and water savings will depend on site-specific designs and climate, and the promised 14-day timeline hinges on fully prepared utilities.