Nvidia's Huang says a three-month AI data center delay costs $10 billion
In brief
- A three-month delay on one large AI data center costs roughly $10 billion in lost revenue, Huang said.
- A 1-gigawatt facility costs about $50 billion and earns about $40 billion a year, according to Huang.
- Nvidia's modular DSX POD can run within 14 days of connecting to existing infrastructure, the company says.
- Nvidia lists 17 rollout partners, including Dell, HPE and Lenovo, with deliveries slated for 2027.
How Huang got to $10 billion
Huang kept the economics simple. A 1-gigawatt AI data center costs about $50 billion to build, and once it's running it's projected to bring in around $40 billion a year in revenue, according to Crypto Briefing's account of his Berlin remarks.
The $10 billion figure comes from the outlet's arithmetic: divide that $40 billion by four, and three months of downtime works out to roughly $10 billion in lost revenue.
That's the number that made the rounds.
Chip supply was the headline worry for a long time. Huang's point was that it isn't the main constraint anymore, and that grid connections and construction timelines are now what's holding projects back.
Nvidia's answer is the DSX POD
Nvidia's proposed fix is the DSX POD, a modular reference design for AI data centers that are built in a factory and shipped in containerized units. Each unit includes Nvidia's next-generation Vera Rubin systems, networking gear and liquid cooling (all integrated at the factory), according to Crypto Briefing.
Huang said the units can be running in as little as 14 days once they're connected to existing infrastructure, compared with about six months under the traditional approach. Nvidia plans a global rollout with 17 partners, including Dell, HPE and Lenovo, and expects deliveries to begin in the first months of 2027.
What the pods don't fix
There's a catch. The 14-day figure only applies after a unit is hooked into existing infrastructure, so it doesn't fix delays in the power hookup itself. Crypto Briefing also noted that Nvidia can't, on its own, speed up the utility and permitting processes that decide when a 1-gigawatt facility actually comes online.
Nvidia also has a lot riding on the buildout. Its 10-Q filing for the period ended July 26, 2026 disclosed $279 billion in supply and capacity commitments, mostly tied to data center infrastructure, as cited by Crypto Briefing.
Frequently asked questions
How did Jensen Huang arrive at the $10 billion figure for a data center delay?
Huang said a 1-gigawatt AI data center costs about $50 billion to build and is projected to generate around $40 billion in annual revenue once it's running. According to Crypto Briefing, dividing that annual revenue by four means three months of downtime equals roughly $10 billion in lost revenue.
What is Nvidia's DSX POD?
The DSX POD is Nvidia's modular reference design for AI data centers that are built in a factory and shipped in containerized units. Each unit includes Nvidia's next-generation Vera Rubin systems, networking gear and liquid cooling, all integrated at the factory. Nvidia plans a global rollout with 17 partners, including Dell, HPE and Lenovo.
Does the DSX POD solve power grid delays for AI data centers?
No. Huang said units can be running in as little as 14 days, but only once they're connected to existing infrastructure. Crypto Briefing noted that this doesn't fix delays in power hookups, and that Nvidia can't, on its own, speed up the utility and permitting processes that decide when a 1-gigawatt facility comes online.
