Nvidia CEO Jensen Huang Declares AGI Has Arrived Following GPT-6 Astra Launch

Nvidia CEO Jensen Huang has declared that artificial general intelligence (AGI) has arrived. The statement follows the launch of OpenAI's GPT-6 Astra, marking a definitive shift in how the industry's most influential hardware provider views the current state of AI.

According to Huang, the arrival of AGI is the result of a rapid four-year evolution, tracing the trajectory from the initial release of ChatGPT through the o1 series and culminating in Astra. The technical foundation for this leap is immense; GPT-6 Astra was trained on a massive compute cluster consisting of more than 100,000 Nvidia Grace Blackwell NVLink72 GPUs. This scale of hardware deployment underscores the sheer volume of processing power now required to push the boundaries of frontier models.

The Shift to Native Multi-Agent AI

The launch of GPT-6 Astra represents more than just a scaling of parameters. The model introduces a shift toward native multi-agent structures, a design that allows the AI to autonomously spawn sub-agents. These agents are capable of navigating software, interacting with web browsers, and managing desktop applications independently, moving the technology closer to a functional digital assistant that can execute complex workflows across different environments.

This development is part of a broader trend of extreme compute scaling. OpenAI has utilized the 'Stargate' project infrastructure located in Abilene, Texas, to support the immense energy and hardware demands of the Astra training run. While OpenAI's official documentation claims GPT-6 Astra achieves a benchmark of 64.6% compared to 52.6% for Claude Fable 5.1, third-party reports on specific coding benchmarks continue to vary.

Industry Implications and the Compute Wall

When the CEO of the primary supplier of AI infrastructure declares that AGI is here, it signals a pivotal moment in industry sentiment. Huang's endorsement validates the current trajectory of "scaling laws," suggesting that increasing compute and data continues to yield transformative capabilities. However, this also highlights an extreme dependency on specialized hardware.

As frontier models move toward multi-agent autonomy, the capital expenditure and energy requirements for the next generation of AI are expected to rise exponentially. The reliance on clusters of over 100,000 GPUs suggests that the barrier to entry for creating AGI-level models is now limited to a handful of organizations with the financial and electrical capacity to maintain such infrastructure.

What Remains Unconfirmed

Despite the bold declaration, several details regarding the scale of the Astra project remain unverified. While the 100,000-GPU cluster is confirmed, specific reports regarding the total financial cost of training—with some estimates ranging between $500 million and $1 billion—have not been officially corroborated. Additionally, while the industry anticipates further scaling, the exact hardware footprint for future training runs remains a subject of speculation rather than confirmed corporate roadmap.