Regarding AI, Huang Renhuang released

On March 18, 2023, local time, Nvidia CEO Jensen Huang unveiled the new AI chip products Blackwell Ultra and GB300 Superchip at the Nvidia GTC 2025 conference.

The GTC 2025 conference was held from March 17 to 21, 2023. More than 1,000 speeches and over 300 on-site demonstrations will be presented at the conference, showcasing breakthroughs in AI applications in cybersecurity, humanoid robots, autonomous driving, medical health, and climate research. In addition, the GTC 2025 conference will also hold China AI Day - Cloud and Internet Online Chinese Special Session, inviting Alibaba Cloud, AutoNation, Meituan, and Galaxy General Robot and other domestic enterprises to share online.

During the keynote speech, Huang unveiled Nvidia's next-generation GPU Blackwell Ultra, which is expected to start shipping in the second half of 2025.

Blackwell Ultra is a product designed for the shift in AI market demand from "training" to "inference", improving accuracy during the inference process. "AI has made great leaps - inference requires higher computational performance by several orders of magnitude," Huang said.

Nvidia introduced the NVIDIA GB300 NVL72 at the conference, which connects 72 Blackwell Ultra GPUs and 36 Grace CPUs in a rack. Its inference speed on large language models is 11 times faster than the first generation of Hopper, with 7 times more computing power.

When using the DeepSeek-R1 model, GB300 NVL72 can process 10 times as many tokens per second as using the Hopper chip. GB300 NVL72 can answer user questions in about 10 seconds, while Hopper takes 1.5 minutes.

Huang said that the new generation GPU Blackwell has been fully mass-produced, "customer demand is incredible," he said, "we will easily transition to the upgraded version (Blackwell Ultra)."

In addition, Nvidia said it will launch its next-generation AI platform Vera Rubin in the second half of 2026 and Vera Rubin Ultra in the second half of 2027.

At the conference, Huang announced that Nvidia has collaborated with General Motors and released the automotive safety AI solution NVIDIA Halos. According to the introduction, Halos covers chips and software, tools, and services, aiming to help automotive autonomous vehicle development safely.

At the same time, Nvidia announced the establishment of a research center in Boston to provide cutting-edge technologies to drive quantum computing development.