A Polygon-based Polymarket contract put the chance of OpenAI formally declaring artificial general intelligence by Dec. 31 at 28%, up from 19% a day earlier and 18 percentage points over the past month. The move followed OpenAI's Sept. 3, 2026 release of GPT-6 Astra and Nvidia CEO Jensen Huang's Sept. 6 X post declaring that AGI had arrived. Huang said Astra was trained on more than 100,000 Nvidia Grace Blackwell NVLink72 GPUs, following the progression from ChatGPT to o1 and Astra over four years. He also said 400,000 additional GPUs were coming online. The assertion revived disagreement over whether advanced task performance and commercially useful AI output meet the threshold for AGI. Nvidia's latest results underscored the scale of AI infrastructure spending behind the claim. Q2 FY2027 revenue reached $96.22 billion, including $89.02 billion from data centers, while the company guided for $108 billion in Q3 revenue. Huang said demand exceeded available supply and emphasized that productive AI tokens turn computing capacity into revenue. OpenAI described Astra as its most intelligent and human-values-aligned model, while President Greg Brockman said the company had entered the AGI era. CEO Sam Altman has called AGI poorly defined, and researcher Gary Marcus said Astra did not meet most of his proposed criteria for the term.