A data centre that arrives on a truck

Nvidia's new DSX Pod puts seven Vera Rubin racks, pre-cabled and liquid-cooled, into a container of about 28 tonnes. The claim is that a site can start serving in 14 days instead of six months. The bottleneck it attacks is not chips but building work.

Tech & Compute Dr. Benjamin Koch / The Desk Wednesday, 21 October 2026, 17:30 CEST 2 min read2 sources

At its GTC conference in Berlin on Wednesday, Nvidia announced the DSX Pod AI Factory: seven of its Vera Rubin AI racks packed into a unit roughly the size of a standard shipping container, weighing about 28,000 kilograms, and designed to travel by truck on normal roads¹. The cabling and liquid cooling are assembled at the factory¹.

What is actually new. Not the chips. The change is where the work is done. “The only thing that has to happen when it gets to the site is it’s placed in its location, connected to the rest of the other containers, and validated through final burning and testing,” Ian Buck, Nvidia’s vice president of hyperscale computing, told reporters¹. Nvidia says this cuts the time for a new AI data centre to start serving tokens from six months to 14 days¹. That figure is the company’s own and is untested until the first pods arrive.

Why Europe gets a mention. Nvidia said the pod is for customers worldwide and could help with deployment problems faced by European customers¹. It did not spell out which. What the design does is move the cabling and cooling work from the site into a factory hall, where it can be done in series and checked before the unit leaves¹.

What it does not change. A pod still has to be placed, connected to the other containers and tested on site¹, and the site still needs its power connection; that does not come on the truck. Nvidia gave no price; analysts estimate a single Vera Rubin rack at more than $7 million¹. Deliveries are expected from early 2027 through manufacturing partners including Foxconn, Gigabyte, Lenovo, HPE, Quanta and Pegatron¹.

The direction of travel. Nvidia has moved from selling single graphics chips to servers, then racks, and now containers of racks¹. Jensen Huang said recently that what Nvidia sells into one gigawatt of data-centre capacity has gone from about $18 billion to “$40-plus billion per gigawatt”¹. The investor question this week, as Meta and Tesla report, is whether buyers keep spending at that rate². The engineering question is simpler and more useful: how much of a building can be made in a factory.

What to watch

A price; Nvidia gave none, and analysts estimate a single Vera Rubin rack at more than $7 million¹.
First deliveries, which Nvidia expects in early 2027 from partners including Foxconn, Lenovo and HPE¹.
Meta's results after Wednesday's US close, and what it says about AI capital spending².

Desk comment

The fastest part of a data centre was always the chip. The slow part was the plumbing, and now the plumbing ships from a factory.

Sources

Nvidia announces plug and play Vera Rubin racks that can be shipped on a truck — CNBC, 21 Oct 2026.
European shares edge up on earnings support; Big Tech results awaited — Investing.com, 21 Oct 2026.

Every figure in this piece traces to a source above or to the desk’s own arithmetic on those sources, marked as such. Quotations are short and attributed. Spot an error? Write to the desk; corrections are logged on the method page.

Commentary for information only. Not investment, legal, tax, medical or political advice. Not an offer or solicitation. Analysis may be incomplete or wrong. Charts show published public data, not a trading feed; binding numbers live with the institutions that publish them.