Economy/Home · Economy

Jensen Huang Says "GPT-6 Astra, AGI Has Arrived"... Reveals Training on 100,000 Nvidia GPUs

Nvidia CEO Jensen Huang, referring to OpenAI's next-generation model GPT-6 Astra, said "AGI has arrived," attaching strong significance to the pace of advancement in artificial intelligence models. CEO Jensen Huang said on his social media on the 7th, "GPT-6 Astra was trained using more than approx

Wooil Shim
Staff Reporter
4 min read
Jensen Huang Says "GPT-6 Astra, AGI Has Arrived"... Reveals Training on 100,000 Nvidia GPUs
CBC News

Nvidia CEO Jensen Huang, referring to OpenAI's next-generation model GPT-6 Astra, said "AGI has arrived," attaching strong significance to the pace of advancement in artificial intelligence models.

CEO Jensen Huang said on his social media on the 7th, "GPT-6 Astra was trained using more than approximately 100,000 Nvidia Grace Blackwell NVLink72 units." He stated, "Four years from ChatGPT to o1 and now Astra. AGI has arrived. Congratulations to the OpenAI team."

He also mentioned that 400,000 additional GPUs are expected to be brought online going forward. This suggests that demand for high-performance GPUs and data center infrastructure is likely to continue expanding in the process of developing large-scale AI models.

The remarks draw attention in that the Nvidia CEO himself disclosed the scale of GPUs used to train GPT-6 Astra and evaluated the model's level of progress in connection with AGI.

However, the statement that "AGI has arrived" is CEO Jensen Huang's personal assessment of Astra. As the definition and achievement criteria for AGI have not been firmly established as a single standard in the AI industry, it is difficult to take this directly as meaning that AGI achievement has been officially confirmed.

If CEO Jensen Huang's remarks are accurate, more than 100,000 Grace Blackwell NVLink72 units were already mobilized for Astra's training, with around 400,000 additional GPUs expected to come online. It also confirms once again that the generative AI competition is expanding beyond model performance to a race to secure infrastructure such as GPUs, HBM, power, and data centers.

[This article was written with AI assistance based on publicly available posts and related materials. Readers are responsible for their own final judgment of quoted statements and figures and their investment decisions.]

Wooil Shim
Staff Reporter

CBC Globe publishes verified stories with editorial review, source checks, and tenant-specific publication standards.