Jensen Huang (젠슨 황), chief executive of Nvidia, said “artificial general intelligence (AGI) has arrived” following the unveiling of OpenAI’s new artificial intelligence (AI) model, Astra.

On Sept. 6, Business Insider reported that Huang posted a message on social media platform X, formerly Twitter, congratulating the OpenAI team and emphasizing that Astra was trained on Nvidia chips.

Huang wrote, “From ChatGPT to o1 to Astra in 4 years,” adding, “AGI has arrived. Congratulations @OpenAI team.” He also wrote, “400K GPUs coming online next,” signaling a further expansion of computing infrastructure. GPUs are key semiconductors used to train and run inference for generative AI models.

OpenAI unveiled Astra on Sept. 4. OpenAI introduced Astra as its most powerful, intelligent and aligned latest model, and said it can handle demanding professional work based on speed, accuracy and judgment. Astra began being offered to some customers on Sept. 4 and was later expanded in stages to paid plans and an API.

OpenAI executives also linked Astra to AGI. Greg Brockman (그렉 브록먼), president of OpenAI, said in a call with reporters, “Welcome to the AGI era.” Brockman said people looking back may think AGI “was made at exactly this time, and probably with this model,” and said he personally believes “we are already at that point.”

In contrast, Sam Altman (샘 알트먼), OpenAI’s CEO, showed a stance somewhat distant from the term AGI itself. Altman recently appeared on the podcast “Sociis” and described AGI as “a very poorly defined term,” adding, “in some ways it is similar to an unimportant marketing term.” The messages around AGI differed even within the same company.

The remarks also tie into a trend in which competition over AI model performance and the semiconductor supply chain are becoming more closely linked. Major AI companies including OpenAI, Meta, Anthropic and Google are developing AI models using Nvidia’s high-performance chips. OpenAI said Astra underwent large-scale training using more than 100,000 GPUs.

That demand was reflected in Nvidia’s results. Nvidia said in its fiscal 2027 second-quarter results announced in August that it posted revenue of $96.2 billion (about 129.4 trillion won), up 106 percent from a year earlier. Data center revenue, which includes AI accelerators, rose 117 percent to $89.0 billion (about 120.0 trillion won).

Huang’s remarks go beyond congratulating OpenAI on the new model, showing that the next round of AI competition requires more computing resources. With Astra’s launch and additional GPUs coming online, competition over AI model performance and the infrastructure that supports it is expected to accelerate together.

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.