Nvidia CEO Jensen Huang declared AGI had arrived with OpenAI’s GPT-6 Astra after a 100,000-plus-GPU training run, though OpenAI has not made that claim.
Key Points:
- Huang said Astra trained on more than 100,000 Nvidia Grace Blackwell NVLink72 GPUs and that another 400,000 GPUs are coming online.
- OpenAI describes Astra as its most intelligent and aligned model, but its Sept. 3 launch announcement does not declare AGI.
- Nvidia’s Data Center revenue reached $89 billion in its latest quarter, up 117% from a year earlier.
Jensen Huang Claim
On Sept. 6, Huang wrote on X that Astra trained on more than 100,000 Nvidia Grace Blackwell NVLink72 GPUs, then declared, “AGI has arrived.” He added that another 400,000 GPUs were coming online.
Huang initially cited 300,000 GPUs in an earlier version of the post, then deleted it and reposted the claim with the lower 100,000-plus figure. Nvidia has not explained the change.
OpenAI introduced Astra on Sept. 3 as its “most intelligent and aligned model,” citing advances in computer use, software engineering, cybersecurity, science and professional work. Media reported Huang’s remarks while noting that OpenAI’s launch announcement did not say the model had achieved AGI. AGI generally refers to software able to handle most intellectual work a person can, though no single definition is universally accepted.
Also Read: iPhone Ultra May Drop Face ID And 4 More Features Despite $2,000+ Price
OpenAI AGI Debate
Huang’s statement carries added weight because Nvidia supplies the computing hardware behind many frontier AI systems, including infrastructure used for OpenAI models. That also gives Nvidia a commercial stake in continued demand for larger AI clusters.
Nvidia reported $89 billion in Data Center revenue for its fiscal second quarter, up 117% from a year earlier. “Now, compute is revenue,” Huang said on the company’s earnings call, arguing that AI demand was accelerating as more labs scaled their models.
OpenAI President Greg Brockman stopped short of making the same categorical AGI claim in the account cited by BeInCrypto. The difference matters because AGI has no universally accepted threshold, leaving companies, researchers and investors to apply different standards to the same technical advances.
The Nvidia-OpenAI relationship has also changed since the companies announced a 2025 plan that envisioned up to $100 billion in Nvidia investment.
That agreement was never finalized, and reports in June said OpenAI was quietly buying fewer Nvidia chips as the company diversified its computing options.
Read Next: Bitcoin Trails Gold By 10X, And CZ Thinks That Ends Next Cycle
