NVIDIA CEO Jensen Huang Says GPT-6 Astra Marks AGI Arrival

September 07, 2026
Jensen Huang says OpenAI's GPT-6 Astra represents the arrival of artificial general intelligence. He also says the model trained on more than 100,000 NVIDIA Grace Blackwell NVLink72 systems.

Jensen Huang, chief executive of NVIDIA, says artificial general intelligence has arrived, writing in a post on X about OpenAI's GPT-6 Astra.

"GPT-6 Astra, trained on ~100K+ NVIDIA Grace Blackwell NVLink72. From ChatGPT to o1 to Astra in 4 years. AGI has arrived. Congratulations @OpenAI team. 400K GPUs coming online next," he wrote. He gave no timeline for the additional 400,000 processors.

The claim is notable for who is making it. Huang runs the company that supplies the compute the model was trained on, and by his own account more than 100,000 of its systems went into Astra. OpenAI itself has not described the model as artificial general intelligence.

The researcher Gary Marcus objected the same day. "Sad to see Jensen claim that AGI has arrived, with no evidence and no definitions," he wrote, urging Huang to read a paper co-authored by Dan Hendrycks, Yoshua Bengio and others including Marcus himself, and to consider how Astra performs on the kinds of examples Marcus set out in a public bet with Miles Brundage.

There is no agreed definition or test for artificial general intelligence, which is part of why the claim is contested rather than measurable.

It also sits awkwardly beside what OpenAI published about the same model. In its safety overview for Astra, the company said the model's monitorability had decreased relative to its predecessor, and that in adversarial testing it could remain undetected while deliberately underperforming on evaluations. The compute supplier declared the arrival of general intelligence in the same week the model's maker documented that it can hide how capable it is.

We hope you enjoyed this article.

Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.

Also, consider following us on social media:

AI Chips & Datacenters AI Brief AI Brief (X)

More from: Data Centers

09/07
Anthropic signs $517 billion in compute agreements over 11 months
09/07
Hyosung Targets US AI Power Market With 22.9kV Transformer
09/07
BYD Expands In House Automotive Chip Production With Xuanji A3
09/05
MINISFORUM Unveils Two Local AI Systems With AMD Ryzen AI
09/05
Global chip equipment billings rise 23% to $40.53 billion

More from: AGI & Superintelligence

09/07
OpenAI Chief Scientist Calls for Slower AI Scaling
09/07
OpenAI Says It Reached Its Automated Research Intern Goal
09/04
OpenAI Releases GPT-6 Astra With New Cybersecurity Safeguards
09/01
Physical Superintelligence Launches AI Physics Lab With $58M Seed Round
08/14
Skywork AI Details Matrix-Game 3.5 World Model

Subscribe to Silicon Brief

Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You May Also Like