5 recipients of the keynote speech by NVIDIA CEO Jensen Huang on Computex 2025
At Computex 2025, NVIDIA CEO Jensen Huang returned to his birthplace in Taipei, Taiwan, to deliver a detailed keynote speech outlining the company’s bold next chapter. Before the Taipei Music Center had a crowd of more than 4,000 people, Huang positioned NVIDIA as GPU Powerhouse, now a full-stack AI infrastructure provider across silicon, systems and software.
His keynote address covers everything from next-generation chip architecture to robotics and supercomputing, which demonstrates Nvidia’s central role in shaping the future of AI across the industry and boundaries.
Here are five of his keynote speeches:
From data center to “AI factory”
Huang called for a conceptual shift in infrastructure that we describe as powering the AI era. He believes that “data centers” are no longer a sufficient term, but rather these facilities should be regarded as “AI factories” in which intelligence is created like any other industrial output.
“You use energy, and it generates incredible value,” Huang said. “These things are called tokens.” Comparing these facilities with past revolutions built on roads and power grids, he stressed that accelerated computing is now an essential national infrastructure. “AI factories are the next great industry and are worth trillions.”
Reshape AI system design through NVLINK fusion
One major technical announcement is the debut of NVLink Fusion, an updated interconnect architecture that extends NVIDIA’s chip ecosystem. The platform will allow third-party CPUs and AI accelerators to be directly linked to NVIDIA GPUs and open the door to semi-regular AI systems through partners such as Qualcomm, Marvell, Marvell, Mediatek and Fujitsu.
Taiwan’s supercomputer
Huang announced the establishment of a partnership between Taiwan’s first AI supercomputer, NVIDIA, Apple supplier Foxconn and the Taiwan government.
The facility aims to strengthen Taiwan’s role as a global AI hub, powered by 10,000 Blackwell GPUs and supports the region’s AI ecosystem, including TSMCs, startups and academic researchers. Starting with Kaohsiung, the building will start with a 20 MW deployment and expand to 100 MW from multiple sites.
Huang pointed out that Nvidia now has 350 partners in Taiwan. “It’s very important to have world-class AI infrastructure in Taiwan,” he said. “But Taiwan is not just building AI for the world; Nvidia is helping to build AI for Taiwan.”
Physical AI systems alleviate labor shortage
Perhaps the most forward-looking part of the yellow keynote, with the focus on robotics and proxy AI
NVIDIA unveiled the Isaac Gr00t N1.5, a fundamental model and development platform that acts as the “AI brain” of humanoid robots, enabling them to reason, adapt and perform complex tasks.
The tool is paired with GR00T-DREAMS, a new simulation environment for training robots in virtual settings designed to solve one of the robot’s main challenges: obtaining high-quality motion data. Huang predicts that these advancements will bring AI-driven manual labor to areas such as logistics, healthcare and manufacturing. “Physical AI and robotics will bring the next industrial revolution,” he said.
DGX Spark: Desktop Supercomputer
In the final reveal, NVIDIA introduced DGX Spark, a compact, high-performance AI supercomputer designed for desktops. Built on GB200 Grace Blackwell Superchip, DGX Spark offers 1,000 tops (trillions of operations per second) to democratize AI development.
By providing this computing power scale outside of a large number of data centers, NVIDIA hopes to empower smaller labs, researchers and even individuals to innovate in AI without the need for enterprise-scale infrastructure.
