NVIDIA's Huang claims artificial general intelligence now exists with OpenAI's latest model

Jensen Huang, chief executive of NVIDIA, announced on September 6, 2026 that the arrival of artificial general intelligence had been achieved through OpenAI's deployment of its GPT-6 Astra system. The infrastructure supporting Astra included over 100,000 units of NVIDIA's Grace Blackwell NVLink72 processors during its development phase, with deployment expected to expand to an additional 400,000 graphics processing units from NVIDIA. OpenAI characterized Astra as representing its most capable and well-aligned iteration to date. However, the pronouncement prompted criticism from Gary Marcus, an NYU-based researcher known for skepticism toward AI claims, who contended that Huang's statement rested on insufficient justification and lacked clarity around the definition of AGI itself, pointing out that the field has yet to establish uniform understanding of what this milestone actually means.

1. Huang declared AGI has arrived on September 6, 2026 regarding OpenAI's GPT-6 Astra release
2. Astra's training relied on deployment of over 100,000 NVIDIA Grace Blackwell NVLink72 processor units
3. An additional 400,000 NVIDIA GPU units are set to come online for Astra
4. Gary Marcus, an AI skeptic and NYU professor, disputed Huang's AGI claim for lacking evidence and definitional rigor