Key Takeaways
- NVIDIA CEO Jensen Huang declared that artificial general intelligence (AGI) has arrived, citing the training of OpenAI’s newest ChatGPT model on his company’s chips.
- Huang highlighted a radical hardware shift: NVIDIA now packs 72 GPUs into a single enclosure that functions as one massive AI brain, a change that required billions of dollars in investment.
- Initial claims mentioned 300,000 of these boxes were used; Huang later revised the figure to about 100,000, without explanation.
- NVIDIA’s AI‑computer sales surged to $89 billion in three months—more than double the prior year—prompting Huang to state, “Now, compute is revenue.”
- OpenAI has not officially confirmed AGI; its president Greg Brockman stopped short of endorsing the claim, and the much‑publicized $100 billion partnership announced for 2025 was never signed.
- Huang closed his social‑media post with a promise to activate another 400,000 chips soon, a figure that analysts say will move markets more than the AGI debate.
Jensen Huang’s Bold AGI Announcement
On a recent Sunday, NVIDIA founder and chief executive Jensen Huang took to social media to proclaim that artificial general intelligence (AGI) has finally been achieved. He wrote that OpenAI’s newest ChatGPT model had been trained on his company’s chips, then added the provocative line: “AGI stands for artificial general intelligence. It means software that can handle most thinking work a person can. Nobody agrees on when that arrives. Huang just called it.” The statement quickly spread across tech circles, sparking both excitement and skepticism about whether the milestone truly marks the arrival of human‑level machine cognition.
The Hardware Leap Behind the Claim
Huang’s confidence rests on a dramatic redesign of NVIDIA’s GPU architecture. Previously, the company sold its accelerators eight to a board; now it integrates 72 GPUs into a single enclosure wired to operate as one colossal brain. “That switch cost billions,” Huang noted in a later interview, “It pays off only if somebody builds something enormous.” The massive scale of this new system is what he believes enabled the training run that produced the latest ChatGPT iteration.
Training Scale: From 300,000 to 100,000 Boxes
In his original post, Huang claimed that roughly 300,000 of the new 72‑GPU boxes had been used to train OpenAI’s model. He subsequently deleted that message and reposted with a lower estimate of about 100,000 boxes. NVIDIA has not clarified why the figure was revised, leaving analysts to wonder whether the adjustment reflects a more accurate count, a correction of an overstatement, or a strategic move to temper expectations. Regardless, the revised number still represents an unprecedented consumption of computing power.
Revenue Explosion: “Compute Is Revenue”
The financial impact of this AI boom is stark. NVIDIA reported selling $89 billion worth of AI‑focused computers in just three months—a figure more than double the same period a year earlier. During the earnings call, Huang distilled the shift succinctly: “Now, compute is revenue.” The remark underscores how the company’s bottom line has become inseparable from the demand for massive AI training infrastructure, turning hardware sales directly into profit as enterprises race to build ever‑larger models.
OpenAI’s Cautious Stance
Despite Huang’s enthusiasm, OpenAI has refrained from endorsing the AGI claim. President Greg Brockman stopped short of affirming that the new ChatGPT model constitutes artificial general intelligence, emphasizing that internal benchmarks still fall short of the broad, human‑like capability the term implies. This hesitation highlights a divergence between NVIDIA’s marketing narrative and the more measured assessment coming from the AI research lab itself.
The Unsigned $100 Billion Partnership
Speculation about a deepened NVIDIA‑OpenAI alliance has been rife, especially after news of a purported $100 billion deal slated for 2025 surfaced. However, BeInCrypto reported in June that the agreement was never signed, and that OpenAI has been quietly reducing its purchases of NVIDIA chips. These details suggest that while the two companies collaborate closely on specific projects, the broader, long‑term financial entanglement Huang alludes to remains speculative rather than contractual.
Future Chip Deployments and Market Focus
Huang concluded his social‑media update with a promise to bring online another 400,000 chips in the near future. He noted that markets will undoubtedly watch that rollout figure closely, as it signals the next wave of compute capacity available for training ever‑larger models. One user reacted to the AGI claim with the comment: “Achieving AGI by 2026 is wild, it was supposed to be 2029+.” The remark captures the tension between Huang’s optimistic timeline and the more conservative expectations held by many in the AI community.
Journalistic Perspective: Hype Versus Verification
As a reporter covering the intersection of hardware breakthroughs and AI progress, it is essential to separate verifiable facts from promotional statements. Huang’s announcement provides concrete data points—namely the revised count of 100,000 72‑GPU boxes and the $89 billion quarterly sales figure—that can be cross‑checked with NVIDIA’s financial filings and supply‑chain reports. Conversely, the AGI declaration lacks an objective, universally accepted metric; no standardized test or benchmark currently exists that definitively confirms artificial general intelligence. OpenAI’s own restraint further underscores the need for caution when interpreting such grandiose claims.
Conclusion: What the Narrative Means for the Industry
Jensen Huang’s proclamation serves as a catalyst for conversation about the pace of AI development, the economics of scaling compute power, and the role of hardware vendors in shaping perceptions of progress. While the numbers he cites—massive chip deployments, soaring revenue, and a revised training scale—are substantiated and impressive, the leap to declaring AGI arrived remains interpretive. Stakeholders—including investors, researchers, and policymakers—should monitor both the tangible rollout of additional chips and any forthcoming, peer‑reviewed evaluations from OpenAI or independent bodies that might clarify whether the newest models truly embody the breadth of human‑like intelligence that AGI promises. Until then, the discourse will continue to oscillate between exhilarating optimism and prudent skepticism.
https://finance.yahoo.com/technology/ai/articles/jensen-huang-says-agi-arrived-213937665.html