Jensen Huang makes his boldest statement yet, considering the completion of training for the Astra model on Nvidia processors to represent the actual arrival point of artificial general intelligence.
Article Index:
- Huang’s controversial announcement on the arrival of general intelligence
- The trajectory of statements and test results of the Astra model
- The issue of conflict of interest and record Nvidia revenues
- A complex relationship and volatile partnerships with OpenAI
- Widespread division in Silicon Valley regarding the truth
- Frequently asked questions
Huang’s controversial announcement on the arrival of general intelligence
Nvidia CEO Jensen Huang announced on Sunday that the era of artificial general intelligence has officially begun. Writing in a striking post on his X account, he stated that OpenAI’s latest model, known as “GPT-6 Astra,” was trained on approximately 100,000 or more Nvidia Grace Blackwell NVLink 72 processors, concluding his post with a definitive phrase: “AGI has arrived,” while promising that 400,000 additional graphics processing units will soon go live.
This direct announcement represents the peak of Huang’s bold statements to date, yet it entails an inherent tension that cannot be ignored, given that Nvidia is the primary commercial supplier of hardware and chips powering the entire artificial intelligence sector and reaping billions of dollars from this accelerating race.
The trajectory of statements and test results of the Astra model
Huang had been paving the way for this historic announcement for months. In March, he told content creator Lex Fridman that he sees general intelligence as now being realized, though he qualified at the time that intelligent agents could not yet found and run a major company like Nvidia. During the August 26 earnings call, he went further, telling financial analysts that AGI has already been achieved for many specialized tasks, describing traditional evaluation metrics and benchmarks as having become obsolete.
Huang tied his claim on Sunday to the record results recorded by the Astra model in advanced tests including FrontierMath Level 4, ARC-AGI 3, and TerminalBench 4.0. Attention was drawn to his initial tweet stating that 300,000 processor boxes trained the model, before he quickly deleted it and republished the lower number without providing any official explanation for the discrepancy.
The issue of conflict of interest and record Nvidia revenues
The irony lies in the fact that OpenAI itself has not officially announced reaching artificial general intelligence, with its officials contenting themselves with describing the model as merely the smartest without crossing that sensitive threshold. This distinction is critically important; AGI, generally understood as a machine’s ability to match human intelligence across a wide spectrum of life activities, lacks a unified agreed-upon scientific definition and has not been confirmed by any major academic or scientific body.
In contrast, Huang sits at the helm of the company selling the machines and engines driving these achievements. In its latest fiscal quarter, Nvidia recorded record revenue of $96.2 billion with 106 percent annual growth, with data center infrastructure accounting for about 92 percent of those staggering sales.
A complex relationship and volatile partnerships with OpenAI
The relationship between Nvidia and OpenAI has experienced volatility and complex positions over the past two years. The massive initial agreement hinted at worth $100 billion in September 2025 was later characterized by Huang in February 2026 as “never a binding commitment,” and Nvidia ultimately injected an investment of only $30 billion in OpenAI’s latest funding round, a figure far below the figures initially circulated.
Nevertheless, mutual dependence remains strong, as OpenAI needs the latest hardware produced by Nvidia’s factories, while the latter needs to impress investors with leading models that justify the continued flow of capital investment toward its advanced platforms.
Widespread division in Silicon Valley regarding the truth
Huang’s position does not enjoy a consensus among tech leaders. Microsoft chief Satya Nadria stated that the industry is “not even close” to achieving general intelligence, while Anthropic chief Dario Amodei estimated the timeframe at between one and three years. Researchers and experts continue to emphasize that current systems still suffer from problems in reliability, long-term planning, and a true understanding of reality.
Huang concluded his post with a future promise indicating the operation of 400,000 additional processors soon, reflecting a clear reality: whether the industry agrees with his controversial definition of AGI or rejects it, chip shipments will continue to flow and sell nonstop.
Frequently asked questions
Question: What did Jensen Huang announce regarding artificial general intelligence?
Answer: He announced the official arrival of artificial general intelligence coinciding with the training of the Astra model on 100,000 Nvidia processors.
Question: Does OpenAI support this sensational claim?
Answer: OpenAI has not announced achieving general intelligence, limiting itself to describing the Astra model as the smartest and most aligned.
Question: How do other tech companies view Huang’s claims?
Answer: There is a wide division, with leaders like Satya Nadella seeing the industry as still far away from this achievement.
