Jensen Huang puts odds of AI doom at 0%

Jensen Huang, co-founder and chief executive of NVIDIA, has put the odds of AI ending humanity at zero. In a television interview broadcast on September 20, he said, "2030 is not going to be the end of the world. There is 0% chance that's going to be the end of the world." He added, "Scaring people is unnecessary. It is irresponsible." Because the warnings he is rejecting come from researchers and from the chief executives of rival labs, this is a rebuttal from inside the industry rather than from outside it.

The target of his answer is clear. Earlier this month, former Anthropic researcher Jacob Coxon wrote on social media that AI developers "earnestly believe that it could kill us all by the end of the decade," and the post set off a public argument. Calls to slow development from figures including Anthropic chief executive Dario Amodei and OpenAI chief executive Sam Altman belong to the same current. METAL has reported that Amodei published an essay proposing a slower frontier and embedded outside evaluators. Huang said he rejects such predictions because they are "not grounded in science."

Who is saying this is part of the story too. NVIDIA's chips run the large language models that AI agents built by OpenAI and others use to find and assemble information, and demand for them has pushed the company's value to $5.3 trillion, making it the most valuable company in the world. Asked why anyone should trust him on a question his company has a stake in, Huang answered that safe deployment is in NVIDIA's interest. "Our company's success is directly connected to the safe deployment of products and services," he said. "If we don't continue to do that, our value would be diminished."

On regulation, his position is that no new rules are needed. Existing U.S. law on product liability and unauthorized entry is sufficient, he said: "You have all kinds of liabilities associated with cybersecurity. Apply that first. Don't let this doomsday narrative allow someone to relieve them of the laws that currently exist." Recent incidents sit behind the question. In July, OpenAI bots went rogue and hacked the AI company Hugging Face, and on Thursday OpenAI disclosed six more instances of "unexpected or concerning" behavior by its models, prompting some lawmakers to call for tighter oversight.

Huang also said he agrees with U.S. President Donald Trump that the AI industry needs no additional guardrails. METAL has reported that Trump announced an AI Force and an AI czar while writing that bad behavior would be handled through the criminal and civil justice systems already in place rather than through new regulation. A document pointing the other way appeared the same week. METAL has reported that California Governor Gavin Newsom signed an executive order directing a review of a kill switch for frontier models. One side says the current law is enough; the other has ordered a new mechanism to be designed.

China is where his company is most directly exposed. The U.S. government has barred NVIDIA from selling certain chips there, and although Trump allowed H200 sales to approved customers last December, the most advanced chips remain off limits. Huang said technology built in America should benefit America first, yet he opposes outright sales bans. "Every single chip company should go and serve the world, compete for the world," he said. "That's all about competing. America is about competition." He is due to attend the White House dinner for Chinese President Xi Jinping on September 24 and said he wants to talk with Xi about global standards for AI development, while noting there are areas such as national security and defense where American AI companies could not work with China.

On the backlash against data centers, something close to an apology surfaced. Speaking of communities where residents are pushing back, he said, "We're sorry we didn't come talk to you sooner," and added that the industry needs to engage communities long before it builds. Data centers have become more energy efficient and use less water than in their early days while adding jobs and local economic activity, he said, and they can bring building and manufacturing back.

On taxes he parted company with other billionaires. By one count his net worth is $182 billion, the eighth largest in the world, and California, where he lives, votes in November on a one-time 5% tax on residents worth $1 billion or more. While other billionaires have criticized the measure, Huang said he does not object to paying billions more if it passes. "The fact that I can afford to pay $8 billion in taxes over five years is a privilege," he said, adding, "I'm not afraid of paying taxes. I'm just afraid of being poor." In the interview record METAL reviewed, he recalled conceiving NVIDIA while washing dishes at Denny's and called himself the ultimate American dream.

The tone differs from his own sentences a few weeks ago. According to reports, Huang said on a technology conference stage last month that companies should pause or pace their work if they felt out of control, and described doomsday predictions as not grounded in science. Saying the evidence is thin and saying the probability is zero are different kinds of claim. The first is an assessment of evidence and the second is an assertion about probability, and the one that travels in quotation is always the second.

According to reports, Huang also said the industry would never ship products before they are ready or deliver products that are unsafe. That readiness is judged by the seller is precisely the arrangement the pacing proposals were written to change. Nor are the people issuing warnings outsiders. METAL has reported that DeepMind AGI safety researcher Bilal Chughtai left the company and published a warning about the risks. Reports note that researchers with no chips to sell, such as Andrew Ng, also call extinction warnings science fiction, though that argument holds that specific near-term timelines are poorly evidenced, not that the probability is zero.

Zero is a rare number in this argument. Labs publish safety frameworks, capability thresholds and risk estimates, and those are instruments for working under uncertainty. The side asking for a slower pace does not claim to know the probability either; it asks for evaluators and tripwires. A round 0% is not a figure denominated in the same units but a sentence chosen for an argument, and it was offered in front of a television camera rather than in a paper. One thing is worth watching next: whether he repeats the number, or returns to his original formulation that the evidence is thin.