I have known Jensen Huang, Nvidia’s CEO, for decades. One thing I learned early is that his public statements often combine engineering ambition, commercial strategy, and Silicon Valley-scale rhetoric.
(Disclosure: Nvidia is among the many global technology companies that subscribe to research from Creative Strategies, the firm I founded)
That context matters when Huang says artificial general intelligence (AGI) has arrived following the release of OpenAI’s GPT-6 Astra. In a social-media post, Huang described the rapid progression “from ChatGPT to o1 to Astra in 4 years” and declared: “AGI has arrived.” Moreover, back in March 2026, he made another assertion about the development of AGI during an episode of the Lex Fridman Podcast, as reported in Forbes, stating that “I think it’s now. I think we’ve achieved AGI.”
Taken literally, however, the statement goes beyond what the public evidence can establish. Under the more demanding definitions used in much of AI research, AGI would be a system that can learn, reason, adapt, and apply knowledge reliably across a very broad range of unfamiliar tasks equal to or above human performance. Astra may represent a major advance, but that is not the same thing as demonstrating true AGI.
Huang’s AGI Claim Uses a Business Definition
Huang appears to be using a practical, economic definition of AGI. In that formulation, AI becomes “general” when it can perform sufficiently valuable intellectual work to help create, operate, or materially support a billion-dollar enterprise.
From the perspective of technology deployment and investment, it is clear why such statements are made. First of all, when the company leader says that AGI has been achieved, it accelerates implementation in businesses and helps justify large investments in the infrastructure necessary for its deployment. Secondly, such a stance is connected to the fact that Nvidia is one of the main enablers of frontier AI training. And last, Huang’s viewpoint reflects the Silicon Valley attitude to AI — when it is already capable of generating substantial economic gains, it has characteristics similar to those of AGI.
Such a viewpoint also reflects the way many firms implement AI technology: not as a single omniscient system, but as a combination of frontier, open, and specialized models which can perform specific tasks. Therefore, for many enterprises, the term “AGI” can mean “AI which is powerful enough to change the industry now.”
Why Critics Still Reject the AGI Label
According to a conventional definition, AGI refers to systems that can learn, reason, and apply knowledge across a broad array of activities at or excelling beyond human levels. Researchers and critics, such as Gary Marcus, claim that current AI models, including Astra, have still failed to show their abilities to reason, plan, understand causal relations, and perform tasks consistently in unfamiliar environments.
Marcus criticized Huang’s stance of declaring victory with “no evidence and no definitions.” Such an opinion means that it is nothing more than corporate rebranding of narrow AI. The fact that there is no single definition of AGI allows such claims but also creates problems for researchers, enterprises, and politicians who should evaluate the risk and capabilities of artificial intelligence.
A Better Term: proto-AGI, not AGI
It is reasonable to claim that the era of proto-AGI has come – an intermediate state of artificial intelligence that is more advanced but not yet…there. The current models demonstrate signs of general intelligence: they can write programs, draft strategic plans, summarize research, and even orchestrate complex workflows. At the same time, modern systems are very fragile, inconsistent, and highly dependent on human guidance. They are good at recognizing and generating patterns but not at understanding, long-horizon planning, or autonomous work in an unpredictable environmen
That does not diminish their importance. It clarifies how they should be deployed. Enterprises should focus less on whether a vendor’s model has crossed an AGI threshold and more on practical questions on productivity benefits, risks, costs and governance.
The More Useful Conclusion
Huang is right about the commercial significance of advanced AI. Systems such as GPT-6 Astra may be powerful enough to create substantial economic value and alter how knowledge work is organized.
That does not resolve the technical question. The evidence shows that AI is becoming more capable, more versatile, and more economically significant but it does not prove that AGI has arrived. The distinction matters because enterprises must plan around the systems available today, not the more autonomous systems suggested by the claims.
This article was originally published on Forbes.com