Nvidia chief executive Jensen Huang faced developers, customers and investors at the company’s annual software developer conference on Tuesday, arguing the company is well-placed to navigate changes in the rapidly evolving artificial intelligence industry.

Speaking at the GTC conference in San Jose, California, which he referred to as the “Super Bowl of AI”, Huang discussed upcoming AI chips, announced “personal AI supercomputers” designed to allow developers to carry out their work on desktops, and spoke about a new platform for humanoid robots.

But for many the focus was on whether Nvidia will be able to sustain the pace after two years of extreme growth as the dominant supplier of accelerator chips for AI data centre tasks.

Compute demand

That growth was called into question in late January after Chinese start-up DeepSeek said it had developed a high-performing AI model with a fraction of the resources expended by large Western tech firms.

Huang said the rise of AI agents will be a driver of continued growth, as the technology requires “easily 100 times more than we thought we needed this time last year”.

He said the top four public cloud vendors, Amazon, Microsoft, Google and Oracle, bought 1.3 million older-generation Hopper AI chips last year, but so far this year have already purchased 3.6 million of the newer Blackwell processors.

Huang also argued that as AI providers shift from training new models to inference, the computational process involved in providing services, Nvidia remains well-positioned as its chips are the only ones that can handle both types of workloads.

Nvidia’s shares closed 3.3 percent lower, indicating investors were not convinced and remain concerned about new tariffs and a possible recession.

Blackwell Ultra

Nvidia said its next GPU, Blackwell Ultra, would roll out in the second half of the year with more memory than the current edition, with a higher-powered platform called Vera Rubin arriving next year followed by Feynman chips the year after.

He said a Blackwell-powered desktop called DGX Workstation would be manufactured by Dell, Lenovo, HP and others, following a smaller desktop launched this year.

The Isaac GR00T N1 platform will “supercharge humanoid robot development”, Huang promised, with Walt Disney and Google’s DeepMind involved in developing it as well as outside developers.

Huang said Nvidia is working with General Motors to use AI in next-generation cars, factories and robots and added that Nvidia is launching a quantum computing lab in Boston.