The popular story goes like this: the world’s biggest tech companies are racing toward one clear finish line called AGI, and whoever crosses it first wins the future. That story is wrong. According to The Verge AI, Nvidia CEO Jensen Huang told investors on Wednesday’s earnings call that the company had already “achieved AGI,” then knocked down his own claim seconds later by calling the whole milestone “senseless.” He’s right about the second part. And that contradiction tells you everything about where the AI industry actually is.
What stands out here is that Huang has done this before. The Verge AI notes he made the same declaration in March on the Lex Fridman podcast, stating flatly, “I think we’ve achieved AGI.” This time he offered no definition and no benchmark. When pressed, he even walked it back, admitting “the odds of 100,000 of those agents building Nvidia is zero percent.” So we’ve achieved human-level general intelligence, but it can’t run the company that builds the chips. Both things can’t be true.
📉 The finish line nobody can locate
The reason Huang can claim victory and dismiss it in the same breath is simple: there is no agreed definition of AGI. None. “Intelligence” itself doesn’t have one either. That’s not a technicality. It’s the whole game.
Look at how the major players describe the same fuzzy idea:
- OpenAI’s charter calls it “highly autonomous systems that outperform humans at most economically valuable work.”
- OpenAI and Microsoft reportedly agreed on a financial version: systems that generate at least $100 billion in profits.
- Anthropic’s Dario Amodei calls AGI “imprecise” and even a “marketing term,” preferring “powerful AI.”
- Meta sells “personal superintelligence,” Microsoft “humanist superintelligence,” Amazon “useful general intelligence.”
- Google DeepMind’s Demis Hassabis says we’ve reached the “foothills of the singularity.”
Five companies, five different flags planted on five different hills. When Sam Altman himself admits AGI is “not a super useful term,” yet says OpenAI will have “something he would call AGI” by year’s end, you’re watching a goalpost get carried, not crossed.
💰 What Huang actually cares about
Here’s the part practitioners should pay attention to. Huang didn’t dwell on AGI because it doesn’t pay the bills. What matters to him is that AI is “doing productive and useful work” and, in his words, “generating profitable tokens.” More compute makes more tokens, more tokens make more profit. “This is the exact phase where we’re at,” he said. “Which is the reason why everybody’s leaning in.”
That’s the honest version of the AI economy right now. The value isn’t in reaching a philosophical milestone. It’s in autonomous agents that learn new skills, improve themselves, and get deployed on real work. Notice that the CEO selling the shovels frames the entire race around token volume, not intelligence. Follow the incentive.
🧭 Why this matters now
AGI talk isn’t harmless marketing noise. It shapes funding rounds, valuations, regulation debates, and hiring. When Ilya Sutskever, who reportedly led OpenAI staff in chants of “feel the AGI,” now runs a company literally named Safe Superintelligence, the word is doing business work, not scientific work.
A few practical takeaways:
- Ignore the AGI headline, track the capability. Ask what a model or agent can reliably do on your actual tasks, not which milestone a CEO announced.
- Treat “we’ve reached AGI” as a hype signal, especially on earnings calls where the audience is investors.
- Watch tokens and unit economics. Huang just told you that’s where the money moves. Cost per useful task beats any grand label.
- Build for agents, not slogans. The real shift, self-improving systems doing repeatable work, is happening whether or not anyone agrees on a name for it.
Expect the AGI claims to keep coming, Huang included, because vague words are excellent hype tools. The smartest move is to stop waiting for a finish line that keeps moving and start measuring what these systems do for you today. You can read the full account at The Verge AI.