NVIDIA to deliver pre-built AI data centres in containers from 2027

NVIDIA is launching DSX POD, a reference design for complete AI data centres that are built inside containers at a factory before they ship. According to NVIDIA, installation at a fully prepared site can drop from six months to as little as 14 days. Partners are due to deliver the first units worldwide in early 2027.

CEO Jensen Huang presented the news during his keynote at GTC Berlin on Wednesday 21 October 2026. NVIDIA Newsroom posted it on X at 11:44 Oslo time, the same day the company published a separate press release. Price, capacity per unit and delivery terms are not disclosed.

What NVIDIA is launching

DSX POD is not a finished product from NVIDIA alone. It is a reference design that partners build to. Each unit holds a complete AI data centre based on NVIDIA's Vera Rubin platform, assembled inside modular containers. Partners complete the rack-to-rack cabling, install the liquid-cooling piping and test everything before the unit leaves the factory.

When the container arrives, it is meant to connect directly to the site's power, cooling, networking and other supporting infrastructure. NVIDIA says this moves the integration and installation of computing, networking, cabling, rack-level power and liquid cooling from the building site to the factory. The company argues that this reduces the planning, construction, specialised labour and logistics needed on site.

The 14-day figure applies to installation once the site is ready. In the keynote, Huang said the timeline applies "once site utilities are ready", meaning power and other utilities are already in place. Securing power, permits and preparing the site are not part of that figure.

Who will build it

NVIDIA says it is working with partners including Bull, Dell Technologies, HPE, Lenovo, Schneider Electric, Vertiv, Foxconn, Eaton, Crusoe and Anduril to deliver DSX POD units worldwide in early 2027. The press release does not say which partners will deliver in the Nordics, and it does not mention Norway.

Huang described the news as a complete transformation of the supply chain. "This will change the way that people install and build data centers," he said, according to NVIDIA's own live blog from the keynote. In the press release, he says an AI factory creates value when it starts producing tokens, not when construction is complete.

What is not disclosed

NVIDIA does not disclose the price, how many GPUs or how much power one unit requires, or what the site requirements are. Nor does it say anything about operations, service and warranty after delivery, or how responsibility is split between NVIDIA, the partner that builds the unit and the site owner. The press release also carries the usual caveat that expectations about performance and availability are forward-looking statements.

What leaders in Norway should note

• Norway has many data-centre operators and companies weighing their own AI capacity. If delivery time really shrinks to weeks, the bottleneck shifts even more clearly to power, grid connection and land. That is where planning should start.
• Pre-built units from one supply chain can mean faster start-up, but also tighter dependence on NVIDIA and its partners' standards. Leaders should ask suppliers about price, service model and who is responsible when something fails.
• First deliveries are announced for early 2027. Investment plans for AI capacity in 2027 can treat modular units like these as a possible alternative to a traditional build, but should not rely on the weeks estimate until it is documented in a concrete offer.

Sources and media

• NVIDIA Newsroom on X: DSX POD announced at GTC Berlin, post published 21.10.2026 at 11:44 Oslo time
• NVIDIA: NVIDIA Introduces DSX POD AI Factories to Accelerate Europe’s Infrastructure Deployment, press release dated 21.10.2026
• NVIDIA blog: GTC Berlin 2026, live updates from Jensen Huang's keynote, live blog dated 21.10.2026
• Sources were checked on 21.10.2026.
• Illustration: AI-generated with Grok Bot / hogby.ai