Jensen Huang delivered a number precise enough to be absurd on CBS News' Sunday Morning: "However it is characterized, 2030 is not going to be the end of the world. There is a zero percent chance it's going to be the end of the world." It was the Nvidia CEO's second public rebuttal of AI doomerism in a week — at the All-In Summit 2026 (September 13-15) he called extinction warnings "fabricated" and "highly irresponsible." According to The Economic Times, in the CBS interview he also backed President Trump's charact
[1][2]erization of AI-risk concerns as a hoax.
In the same week, Anthropic CEO Dario Amodei published a roughly 3,800-word essay urging the industry to slow its release cadence and give safety evaluations time to catch up; Musk, Altman and Hassabis publicly agreed. The four labs were then sued in the Northern District of California over the coordinated slowdown (Buist v. Anthropic PBC, 3:26-cv-10693). "Those who make AI are braking; those who sell chips are flooring it" — a line used to summarize the week. But it is worth separating the brake from the accelerator and asking which balance sheet each is attached to.
Start with incentives. Amodei's Anthropic is a net buyer of compute, purchasing on a scale of tens of billions of dollars; Huang's Nvidia is the net seller, with the overwhelming majority of revenue from data-center chips. A slowdown is a safety premium for the buyer — capabilities grow slower, evaluations keep pace — and a demand risk for the seller: slower training and inference means slower order growth. None of this proves either position false; it explains why Huang treats "risk concern" as something to rebut rather than manage. His rebuttals also have grounding — at All-In he cited failed predictions like "AI will fully take over radiology within five years" — but past misses do not generalize into "the next prediction will miss too."
The two diverge not only on speed but on who should govern AI. Huang publicly endorsed Trump's "hoax" framing; Trump, on the same day, announced an "AI Force" task force and a search for an "AI czar," signaling for the first time stronger federal oversight — of the loosening kind. Amodei's three-step plan runs the other way: independent evaluators inside labs, shared safety standards and capability-growth caps among democratic countries' AI companies, and eventually global coordination. One wants government power to keep AI from being slowed by risk narratives; the other wants industry self-restraint to slow it down deliberately. The two companies give opposite definitions of regulation.
Put the positions back into the capital markets and the irony sharpens: Anthropic is racing toward an IPO valued near $2 trillion, and Amodei's slowdown essay landed almost in sync with the roadshow calendar; Nvidia maintains its seller's market in AI chips while "0%" underwrites the long-term demand outlook. Trump's surprise phone call into Huang's All-In appearance endorsed the "hoax" framing — politics and the supply chain locked together in one week.
Between "0%" and "slow down" there is no technical disagreement, only a disagreement about timelines: one company profits from selling compute, the other competes by buying it. The public argument about AI in 2026 increasingly looks like two balance sheets talking — and the sentence that sounds most like "objective fact" is usually the one closest to the speaker's own ledger.
[1][2]