Nvidia unveils modular AI factory pods, aiming to cut data centre build time to 14 days

DSX Pod system announced at Berlin GTC conference; Australian neocloud Firmus appears as partner amid funding challenges

By LineZotpaper AI staff

Published28 October 2026

Updated28 October 2026

Read Time3 min

Follow this storyHear when it moves

Illustration: Nvidia unveils modular AI factory pods, aiming to cut data centre build time to 14 days

Nvidia has unveiled a modular "pod" system for AI data centres that it says can reduce deployment time from six months to as little as 14 days. The DSX Pod, announced at the GTC conference in Berlin, consists of prefabricated units the size of shipping containers that arrive at a prepared site ready to connect to power, cooling and networking infrastructure.

Nvidia aims to ship modular artificial intelligence factory "pods" packed in a box about the size of a standard shipping container, as it bids to reduce the time needed to set up AI data centres to as little as two weeks.

The pods "will change the way people build AI centres," Nvidia chief executive Jensen Huang told attendees of the Nvidia GTC conference in Berlin.

The announcement of what Nvidia calls the DSX Pod focuses on mitigating delays in the construction of data centres and reducing the time it takes AI systems to process.

"Each gigawatt data centre generates tens of billions of dollars (revenue) per year, and so if we just said it's $40bn per year of revenue generation that comes out of that infrastructure, if that is delayed by just three months that's $10bn worth of revenues," Huang said.

"DSX Pod allows AI factories to be built at the speed of light" meaning "more intelligence per watt (and a) shorter time to first token."

Nvidia said in a statement: "The units arrive ready to connect out of the box to a prepared site's power, cooling, networking and other supporting infrastructure, reducing deployment time from six months to as little as 14 days."

Huang said: "An AI factory creates value when it starts producing tokens – not when construction is complete. We're bringing factory-built precision and speed to AI infrastructure construction reducing deployment time from months to weeks so customers can scale capacity and bring revenue-generating production online faster."

The pods will be built by Nvidia partners including Armada, Bull, Crusoe, DayOne, Schneider Electric, Lenovo, Dell Technologies and Wistron. They are slated for rollout in early 2027. Nvidia partners will complete rack-to-rack cabling, wiring, piping and functional validation before shipment.

Separately, the conference featured Australian neocloud Firmus, in which Nvidia has an equity stake and to which Nvidia supplies AI chips. Firmus co-founder Oliver Curtis was in Berlin shoring up a funding deal with Nvidia after the collapse of the AI factory company's ASX float, according to The Australian. Firmus has a contract with ASX-listed Maas Group for Maas subsidiary JLE to deliver its modular Power Cubes for AI factory projects, and claims its technology uses 33% less energy and 99% less water to cool GPUs. The Australian earlier this week reported that Firmus is seeking an emergency $US3bn private raising after its planned $44bn sharemarket listing in Australia fizzled when investors baulked at an outsized valuation and limited track record.

Sydney-headquartered Iren, UK-based NScale, Dutch Nebius and US Lambda were also featured as AI factory partners in Nvidia's product launch video.

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

Analysis

Why This Matters

Rapid deployment of AI data centres is critical as demand for computing power surges; delays of months can mean billions in lost revenue for operators.
Nvidia's pod approach could accelerate the build-out of AI infrastructure globally, affecting cloud providers, startups and energy consumption patterns.
The rollout in early 2027 sets a timeline for when large-scale prefabricated AI factories become commercially available, potentially reshaping competition among data centre builders.

Background

Large AI data centres typically take six months to a year to construct, with site preparation, cabling and cooling setup causing most delays. Nvidia, as the dominant supplier of AI chips, has pushed to control more of the infrastructure stack. The DSX Pod is its latest attempt to standardise and speed up deployment, similar to modular data centre concepts used by other tech firms but tailored to high-density GPU workloads.

What to Watch

Whether Nvidia's partners can deliver on the 14-day deployment claim once pod manufacturing begins in 2027.
The outcome of Firmus's emergency capital raise and its ability to fulfill contracts with Meta and OpenAI.
How competitors such as AMD and cloud providers like Amazon Web Services respond to the DSX Pod approach.

Sources

Nvidia unveils modular AI factory pods to build data centres in 14 days · Stockhead

Checked

12 claims in this story checked against 1 source: 9 supported.

Show each claim

Partly supportedNvidia is aiming to cut data-centre build time to 14 days.Stockhead: “reducing deployment time from six months to as little as 14 days”
Not in the sourcesThe summary describes Firmus as an Nvidia partner.Stockhead: “Firmus was briefly displayed as a neocloud on the big screen behind the billionaire during his speech, which listed a large group of companies Nvidia deals with across the AI stack.”
SupportedNvidia says the DSX Pod can reduce deployment time from six months to as little as 14 days.Stockhead: “The units arrive ready to connect out of the box to a prepared site’s power, cooling, networking and other supporting infrastructure, reducing deployment time from six months to as little as 14 days”
Partly supportedThe DSX Pod was announced at the GTC conference in Berlin.Stockhead: “Separately in Berlin, Nvidia also unveiled its DSX AI factory platform”
SupportedThe pods are packed in boxes about the size of standard shipping containers.Stockhead: “Nvidia aims to ship modular artificial intelligence factory “pods” packed in a box about the size of a standard shipping container”
SupportedHuang told GTC attendees that the pods “will change the way people build AI centres”.Stockhead: “The pods “will change the way people build AI centres,” Nvidia’s billionaire chief executive, Jensen Huang, told attendees of the Nvidia GTC conference in Berlin.”
SupportedHuang said a three-month delay to a gigawatt data centre could mean $10bn in lost revenue.Stockhead: ““Each gigawatt data centre generates tens of billions of dollars (revenue) per year, and so if we just said it’s $40bn per year of revenue generation that comes out of that infrastructure, if that is delayed by just three months that’s $10bn worth of revenues,” Huang said in Berlin.”
SupportedHuang said DSX Pod allows AI factories to be built “at the speed of light”, meaning “more intelligence per watt” and a “shorter time to first token”.Stockhead: ““DSX Pod allows AI factories to be built at the speed of light” meaning “more intelligence per watt (and a) shorter time to first token.””
SupportedHuang said an AI factory creates value when it starts producing tokens, not when construction is complete.Stockhead: “Huang said: “An AI factory creates value when it starts producing tokens – not when construction is complete.”
SupportedFirmus has an Nvidia equity stake and Nvidia supplies it with AI chips.Stockhead: “Nvidia has an equity stake in Firmus and supplies it with its AI chips.”
SupportedFirmus claims its technology uses 33% less energy and 99% less water to cool GPUs.Stockhead: “what it claims to be its own technology that uses 33% less energy and 99% less water to keep high-performance graphics processing units, or chips, cool in its AI factories.”
SupportedThe Australian reported that Firmus was seeking an emergency US$3bn private raising after its planned $44bn Australian listing fizzled when investors baulked at its valuation and track record.Stockhead: “The Australian earlier this week revealed that Curtis had flown in his private jet from Singapore to Berlin, and that it is seeking an emergency $US3bn private raising after its hyped $44bn sharemarket listing in Australia fizzled when investors baulked at an outsized valuation and limited track record.”

Correction

28 October 2026An earlier version of this story said Firmus supplies Nvidia with AI chips. Nvidia supplies Firmus with AI chips.