Nvidia's Huang Puts Agent Software at About $2,000 per GPU a Year; GTC Berlin Also Brought DSX POD and OpenShell

Updated October 26, 2026: Added Jensen Huang's remarks at GTC Berlin on what Nvidia charges for its agent software, as reported by Tom's Hardware. The original October 21 report is below, unchanged.

October 26: Huang on software pricing#

"Software licensing will become increasingly a major part of our company's business," Huang said at GTC Berlin, according to Tom's Hardware, which published his remarks on Sunday, October 25. "The way that we license will be…for example, it's about $2,000 per GPU per year to run all the software that Nvidia provides that runs agents on these AI factories," he said.

Tom's Hardware notes that Nvidia's AI Enterprise platform is listed at $4,500 per GPU for a one-year subscription, and that Nvidia has not confirmed whether the two prices cover the same products or terms. It says the lower figure "should not be treated as a confirmed price cut." Nvidia has not said when the roughly $2,000 fee would apply or which software it would cover, the outlet reports.

Limits of the October 26 update#

• Huang framed the $2,000 figure as an example; we treat it as an illustration, not a published price list.
• This rests on Tom's Hardware's account of a video of Huang's comments; we did not find a second outlet or an Nvidia statement on the pricing.
• Tom's Hardware dates Huang's remarks to October 22, while our original report and Nvidia's releases date the GTC Berlin announcements October 21.
• We have left out revenue projections built on the $2,000 figure, which come from an analyst's post quoted by Tom's Hardware, not from Nvidia.

Nvidia is selling further up the stack at both ends: the whole data hall as a shippable product, and the guardrails for the agents that run in it. At GTC Berlin on Wednesday, October 21, Nvidia introduced DSX POD, a reference design for complete Vera Rubin AI factories that partners build inside modular containers and validate before delivery, and launched the OpenShell Reference Architecture for containing and monitoring AI agents.

DSX POD: the AI factory as a container#

Nvidia says the units arrive ready to connect to a prepared site's power, cooling and networking, reducing deployment time "from six months to as little as 14 days." Partners do the cabling, wiring, piping and functional validation before shipment, it says. CNBC reports that each POD holds seven Vera Rubin racks, is roughly the size of a standard shipping container and weighs about 28,000 kilograms, according to Nvidia. "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 reporters, per CNBC.

Nvidia did not give a price. CNBC says analysts estimate a single Vera Rubin rack can cost more than $7 million. Nvidia says partners including Bull, Dell Technologies, Foxconn (Ingrasys), HPE, Lenovo, Schneider Electric and Vertiv will deliver DSX POD AI factories worldwide in early 2027; CNBC says Nvidia pitched the design partly at deployment challenges faced by European customers.

OpenShell: agent containment in hardware#

The OpenShell Reference Architecture combines Vera Rubin systems for model inference, Vera CPUs for agent harnesses and tool use, and BlueField-4 DPUs running Nvidia's Sentry security software, built on Nvidia's OpenShell secure runtime. Nvidia says that if an agent tries to move outside its software boundary, Sentry "quarantines and stops it in milliseconds," and that Sentry on BlueField-4 boosts agent capacity by up to 20%. "Agents can think on Rubin, act on Vera and be governed on BlueField," said CEO Jensen Huang.

Nvidia says the architecture is available now. It lists Anthropic, AWS, Google Cloud, Meta, Microsoft and OpenAI among the companies working with it. Nvidia describes them as working with the architecture and does not say which have deployed it. We covered Nvidia's earlier plan to use OpenShell and Sentry in UK agentic-safety research.

Limits#

• The 14-day deployment figure, Sentry's response time and the capacity gain are Nvidia's own claims; the 14 days assumes a site whose power and cooling are already prepared.
• The rack price is an analyst estimate reported by CNBC, not an Nvidia figure.
• DSX POD deliveries are scheduled for early 2027.

Sources#

• NVIDIA Newsroom: NVIDIA Introduces DSX POD AI Factories to Accelerate Europe's Infrastructure Deployment (October 21, 2026)
• CNBC: Nvidia announces plug and play Vera Rubin racks that can be shipped on a truck (October 21, 2026)
• NVIDIA Newsroom: NVIDIA Launches OpenShell Reference Architecture for Safe AI Agents (October 21, 2026)
• Times of AI: NVIDIA Pledges Compute Valued at $250 Million Over Five Years for UK Agentic AI Safety Research (October 14, 2026)
• Tom's Hardware: Nvidia discusses $2,000-per-GPU annual fee for software that 'runs agents on AI factories' (October 25, 2026)