Jensen Huang Reveals 400,000 Additional GPUs for Training, Mentions GPT-6 Astraeus

Nvidia CEO Jensen Huang celebrated the launch of OpenAI's GPT-6 Astraeus by stating that the model was trained on more than 100,000 Nvidia Grace Blackwell NVLink72 GPUs. Huang shared on X that it took four years from ChatGPT to Ohwon and then to Astraeus. He declared 'The AGI era has arrived,' though this statement is considered a personal opinion without confirmation as a technical standard. OpenAI previously stated in its September 3, 2026, safety overview for GPT-6 Astraeus that it is the highest-performing model among those widely deployed by the company. Huang also mentioned plans to sequentially bring an additional 400,000 GPUs online, though deployment locations and specific activation dates remain unconfirmed.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 40만 GPU 추가 가동 계획, 젠슨 황이 언급