Nvidia CEO Claims AGI Arrival | What OpenAI Astra Means

Jensen Huang Declares the Arrival of AGI

Nvidia CEO Jensen Huang publicly asserted that artificial general intelligence has arrived. He made this claim on Sunday following the release of Astra, the latest model from OpenAI. Huang used his platform on X to congratulate the team behind the product. He highlighted the rapid progression of OpenAI technology, noting a four-year window that spanned from the original ChatGPT to the recent launch of o1 and now Astra. By confirming that the model trained on Nvidia hardware, Huang linked his company’s infrastructure directly to this purported technological milestone.

OpenAI introduced Astra last Thursday, describing it as its most intelligent and aligned model to date. The company claims the system manages demanding professional tasks with high levels of speed and accuracy. President Greg Brockman signaled the transition on a call with reporters, stating that the company has entered the AGI era. OpenAI defines AGI as an autonomous system that outperforms humans at most economically valuable work. The product is currently rolling out to users.

The Debate Over Definitions

Not everyone in the technology sector accepts the claim that AGI exists. Critics argue that the term remains poorly defined and is often used for marketing purposes. Gary Marcus, a prominent researcher, responded to Huang’s statement by noting a lack of evidence. Marcus pointed out that Astra fails to meet many standard academic benchmarks for general intelligence. He described the announcement as an attempt to control a scientific discussion through corporate influence.

Others hold similar reservations. The researchers behind the ARC Prize acknowledged that Astra shows significant improvement but cautioned that success on their specific tests does not equate to AGI. They noted that these benchmarks operate in constrained environments. Real-world conditions involve a level of complexity and open-endedness that current models cannot replicate. Even Sam Altman has previously expressed doubt about the term, calling it an imprecise label that carries little weight.

Infrastructure and Future Scale

Nvidia remains the central provider of the compute power required for these models. OpenAI explicitly identified Nvidia as the foundation of its infrastructure in a March funding statement. The hardware dependency is clear, as the vast majority of OpenAI’s inference and training tasks run on Nvidia GPUs. This reliance has driven massive financial returns for the chip manufacturer, which reported $96.2 billion in quarterly revenue this past August.

The scale of this hardware buildup appears set to continue. In his post on Sunday, Huang mentioned that 400,000 additional GPUs are coming online soon. This expansion suggests that while experts debate the definition of intelligence, the industry focus remains on deploying larger clusters of processors. Whether these machines reach the threshold of human-level reasoning or simply become more efficient at specific tasks, the current trajectory points toward continued hardware integration.