Glonce

Jensen Huang says Nvidia achieved AGI again, calls it senseless

miscThe Glonce editorsThe Verge🔥47Glonce rating47Impact25Novelty20Relevance100Surprise15Players85Credibility85
Jensen Huang says Nvidia achieved AGI again, calls it senseless

On Nvidia's earnings call Wednesday, CEO Jensen Huang said that 'for many tasks, we could say that we've already achieved AGI.' He gave no precise definition or benchmark for the claim. In the same remarks, he dismissed the idea of AGI as a meaningful milestone at all: 'I think of all of those milestones and all those, you know, they're kind of senseless at this point.' Huang pointed to AI's shift from answering simple prompts to autonomous agents that can learn new skills and improve themselves 'recursively.' What actually matters, he said, is that AI is 'doing productive and useful work' and 'generating profitable tokens,' with more compute producing more tokens and, in turn, more profit: 'This is the exact phase where we're at. Which is the reason why everybody's leaning in.'

It was not the first time Huang has said AGI has been reached. In March, on the Lex Fridman podcast, he stated plainly, 'I think we've achieved AGI,' without specifying what he meant. Fridman proposed his own, separate definition on the spot: an AI system able to 'essentially do your job,' meaning start, grow and run a successful tech company worth more than $1 billion. Huang later walked the claim back: 'the odds of 100,000 of those agents building Nvidia is zero percent.'

The wider AI industry has settled on no shared meaning for AGI either. OpenAI's charter defines AGI as 'highly autonomous systems that outperform humans at most economically valuable work,' but OpenAI's Altman has admitted that AGI is 'not a super useful term.' Complicating that further, OpenAI reportedly worked out a separate, profit-based definition of AGI with Microsoft: systems able to generate at least $100 billion in profit. In a recent Time story, OpenAI chief research officer Mark Chen put the company's own progress at '80% of the way' to AGI, and Altman has said OpenAI would have something he'd call AGI by the end of the year.

Other companies use their own, similarly loose language for the same idea. Anthropic CEO Dario Amodei calls AGI 'imprecise,' even a 'marketing term,' and prefers to talk about 'powerful AI' instead. Meta talks about 'personal superintelligence,' Microsoft about 'humanist superintelligence,' and Amazon about 'useful general intelligence.' Google DeepMind's Demis Hassabis describes the current moment as the 'foothills of the singularity,' and OpenAI cofounder Ilya Sutskever, who reportedly led employees in chants of 'feel the AGI,' now runs a company called Safe Superintelligence. None of these companies are reported as giving a fresh comment for this story; their terms are cited as existing usage, not new reactions to Huang's remarks.

Key facts

  • On Nvidia's Wednesday earnings call, CEO Jensen Huang said 'for many tasks, we could say that we've already achieved AGI' without giving a precise definition or benchmark, then called such milestones 'senseless.'
  • It wasn't his first such claim: in March, on the Lex Fridman podcast, Huang said 'I think we've achieved AGI'; Fridman proposed his own definition (an AI able to start, grow and run a company worth more than $1 billion), and Huang later walked the claim back, saying 'the odds of 100,000 of those agents building Nvidia is zero percent.'
  • OpenAI's charter defines AGI as 'highly autonomous systems that outperform humans at most economically valuable work,' but OpenAI reportedly worked out a separate, profit-based definition with Microsoft: at least $100 billion in profit.
  • Per a Time story cited in the piece, OpenAI chief research officer Mark Chen estimated the company is '80% of the way' to AGI, and Altman has said OpenAI would have something he'd call AGI by the end of the year.
  • Other companies use their own terms for similar ideas: Anthropic's Dario Amodei calls AGI 'imprecise' and a 'marketing term,' preferring 'powerful AI'; Meta says 'personal superintelligence,' Microsoft 'humanist superintelligence,' Amazon 'useful general intelligence,' and Google DeepMind's Demis Hassabis describes the 'foothills of the singularity.'

Why it matters

AGI is treated across the industry as the AI race's finish line, and this story is a clean case of why that finish line keeps moving. Huang says Nvidia has 'already achieved AGI' for many tasks, offers no benchmark for the claim, and in the same remarks calls AGI milestones 'senseless' because there's no agreed definition of what one is. He ties what he says actually matters, that AI is 'doing productive and useful work' and 'generating profitable tokens,' directly to more compute producing more tokens and, in his words, more profit. The article traces the same pattern through the rest of the industry: OpenAI's charter sets a capability bar ('highly autonomous systems that outperform humans at most economically valuable work'), OpenAI and Microsoft reportedly worked out a separate profit-based bar of at least $100 billion, and OpenAI's own chief research officer and Altman each add a looser sense of the term: Mark Chen's '80% of the way' estimate, and Altman's own prediction that OpenAI would have something he'd call AGI by the end of the year. When bars this different can each be called 'AGI' by the companies chasing it, the article's own conclusion is that the term is senseless unless, as it puts it, you want 'a handy tool for hyping up progress.'

Who it affects

Anyone reading 'AGI achieved' as a marker of real technical progress, including Nvidia's own customers and investors, who hear Huang connect the claim directly to compute demand and profit. It affects OpenAI most directly: the company was founded with AGI as its explicit goal, and the article lays out four different ways it or its leaders have framed the term, its charter's capability definition, the reported $100 billion profit threshold worked out with Microsoft, chief research officer Mark Chen's personal '80% of the way' estimate, and Altman's prediction of something AGI-like by the end of the year, none of them the same measure. It also affects how claims from other companies should be read side by side: Anthropic's Dario Amodei, Meta, Microsoft, Amazon and Google DeepMind's Demis Hassabis each use their own separate term for a broadly similar idea, 'powerful AI,' 'personal superintelligence,' 'humanist superintelligence,' 'useful general intelligence' and the 'foothills of the singularity,' none of them measured against a shared bar either.

How to use it

There's no product to use here, but the piece hands readers a filter for the next 'AGI achieved' headline. Check whether a specific benchmark or definition comes with the claim: Huang's earnings-call statement did not have one, by the article's own account, and neither did his March claim on Lex Fridman's podcast. Keep a claim's origin straight: the $1 billion company threshold is Fridman's own proposed definition, not something Huang endorsed or was measured against. Treat hedged reporting as hedged, not confirmed: the article introduces both the $100 billion OpenAI-Microsoft profit figure and Sutskever's 'feel the AGI' chants as reported, not confirmed by the companies involved. And separate a personal estimate from a measured result: Mark Chen's '80% of the way' is his own estimate, given to Time, not a benchmarked score.

How solid is it

The central claims come from Huang's own remarks, made on Nvidia's earnings call and, for the March claim, on the Lex Fridman podcast. The article itself notes he gave no precise definition or benchmark for either claim, and it treats his own contradiction, an unqualified 'we've achieved AGI' in March against 'zero percent' odds for the same underlying scenario later in that same podcast appearance, as evidence the claims don't hold up even by Huang's own account. Two of the piece's most eye-catching figures carry an explicit hedge: the $100 billion OpenAI-Microsoft profit threshold and Sutskever's chants of 'feel the AGI' are both flagged as reported, not confirmed by the companies named. Mark Chen's '80% of the way' comes from a separate Time story and is presented as his personal estimate rather than a measured result. None of the companies named, OpenAI, Microsoft, Meta, Amazon, Anthropic or Google DeepMind, are reported as giving The Verge a fresh comment for this piece; every term attributed to them is characterized as existing usage the article is recounting, not a new reaction to Huang's remarks.

Risks and caveats

Because AGI has no agreed definition or measurable bar, a claim that it has been 'achieved' can be made, dismissed as 'senseless' and partly walked back, as Huang did between March and this earnings call, without any external check to hold it against. The $100 billion profit figure and Sutskever's chants should not be read as confirmed: the article introduces both as reported, not verified by the companies involved, and Mark Chen's '80% of the way' as his personal estimate rather than a measured score. The piece does not report Nvidia's actual financial results, revenue, profit or guidance, from the same earnings call; it covers only Huang's AGI remarks made during it. No calendar year is attached to any of the article's relative dates, 'Wednesday,' 'in March,' 'last year,' 'by the end of the year,' so exactly when each falls is not stated.

“I think of all of those milestones and all those, you know, they're kind of senseless at this point.”

— Jensen Huang, Nvidia CEO