Jensen Huang

Huang expects the cost of intelligence to fall quickly as more compute comes online and models become more efficient, with demand more than offsetting it.

Um in the in the coming years, you know, my expectation is is that the the cost of intelligence would fall fairly fairly quickly,

Huang says his $100 billion physical AI market estimate was understated, as NVIDIA already did $10 billion in physical AI in the first half of the year.

I I understated the hundred billion dollars because we're we're already the first six months of this year we did $10 billion in physical AI.

Huang predicts that industries touching the physical world, starting with automakers, will all become technology and AI companies.

I think every car company is going to be a technology company.

Huang argues perfect alignment is unattainable, so AI safety must rely on zero-trust containment and out-of-band monitoring instead.

alignment is going to be a forever problem. >> Yes. >> If we depend on perfect alignment in order to ensure safety, then we're never going to get there.

Huang says AI safety and security are inseparable from capability and must be accelerated together.

When I say we need to accelerate AI's technology, what I mean by that is we need to accelerate the security technology. We have to accelerate the safety technology. They all have to be accelerated.

Huang calls safe and secure AI development an existential issue for the AI industry itself.

there will be no AI technology industry if we don't build it safely and securely okay this is an existential issue for us

Huang says demand exceeds the projected 43GW buildout, but land, power and data center shells will be the key bottleneck next year.

But I do think that land powered shell is going to be a challenging part next year.

Huang says power scarcity pushes customers toward NVIDIA's highest tokens-per-watt platform, so its market share is increasing.

But surprisingly to everybody, not only are we a large company and our market share is high, our market share is actually increasing for these reasons.

Huang says NVIDIA's content per gigawatt has risen from about $18 billion to over $40 billion and will keep rising as it adds networking, CPUs and security.

the offering that we we offer into one gigawatt went from went from uh uh I guess about you know 18 billion or so per gigawatt to 20 somewhat billion per gigawatt to now 40 plus per gigawatt

Huang says NVIDIA is becoming one of the largest and fastest-growing CPU companies.

We're now going to be one of the largest CPU companies, surely the fastest growing one.

Huang says NVIDIA and SpaceX are co-developing space data center architecture and that NVIDIA's high-value compute justifies launch costs.

And so we're the perfect fit really to launch into space right now. Irrespective of the com the launch cost, it'll still be excellent ROI.

Huang argues the market puts too much pressure on OpenAI and Anthropic, since frontier labs are only about a quarter of NVIDIA's compute and the buildout doesn't depend on them alone.

But they don't have to be anywhere near a trillion dollars in revenues to be successful.

Huang says with Moore's law over, compute spending must scale with demand, driving semiconductors toward becoming the largest industry.

But if you want twice as many transistors now, doesn't matter how long you wait, you have to pay twice as much.

Huang says annual AI infrastructure spending of $3-4 trillion by decade end could keep growing toward $10 trillion.

runs around 10 trillion dollars. I'm not even a little bit surprised.

Huang advocates open models, saying 80% of AI startups use them to build domain-specific superintelligence rather than compete with frontier labs.

And if you look at all these startup companies, 80% of them use open models.

Huang says substrates, silicon photonics and other inputs are supply constrained and the demand-supply gap will be hard to close for years.

silicon photonics is is supply constrained and so so everything is supply constrained

Huang explains NVIDIA gives all its business to TSMC because only it offers five-year planning visibility, capacity and agility.

The reason why we give 100% of our business to TSMC is because they're the only company in the world that I can plan five years out,

Huang argues general-purpose GPUs beat custom chips because fungibility and durability lower risk and TCO for huge infrastructure bets.

So specialization is the enemy of fungibility. >> Specialization is the enemy of durability.

Huang says enterprise software licensing, at a few thousand dollars per GPU per year, will become a very large NVIDIA business.

And so software license for enterprise business will be very very large.

Huang says NVIDIA will likely move to revenue-share deals with neoclouds it supports with offtake, software and financing.

And so in in those kind of relationships, it's very likely that we'll move to Refshare. And so we'll make an investment. We'll get the benefit of the sale up front, but we'll also have the benefit of Refshare going forward.

Huang says NVIDIA's investments as a share of cash generation will fall quickly since opportunities like OpenAI and Anthropic won't recur.

as a percentage of of our total um uh cash generation it's going to decline very rapidly partly because open AI and anthropic isn't going to come around again

Huang rejects AI doomerism, saying he is 99.99% confident the outcome will be good because many people are working responsibly.

And we should express our worries. We should express our concerns. But we shouldn't predict the end of the world.

Huang expressed optimism about CoreWeave, saying its strength is improving daily and its cost of capital is falling.

CoreWeave's strength is improving every day," adding, "I'm optimistic. Their cost of capital is actually decreasing,

Jensen Huang says accelerated computing has turned cars into intelligent machines and NVIDIA will combine Hyperion and Halos with Volvo's safety systems to build safer next-generation cars.

Accelerated computing has transformed the car into an intelligent machine that can understand the world around it. Volvo has spent nearly a century protecting people on the road. Together, we’re combining Volvo’s extraordinary vehicle safety systems with NVIDIA Hyperion and Halos to build a new generation of cars that make driving safer and more effortless.

Huang criticized the UK for overthinking and finding reasons not to act despite strong talent and culture.

"You have smart people, you have a great culture, you have young people, and the energy is vibrant. But for some reason, if there was a million reasons to win, you find the one reason not to do it."

Huang said the US and Chinese AI sectors must cooperate to monitor risks of potentially dangerous AI and advance it safely.

We absolutely should work with each other so that we advance the technology safely

Huang called for enhanced US-China AI dialogue, noting he already holds dialogues with many researchers in China and has urged the UK to do the same.

We ought to enhance the dialogue,

Huang argues that x86 hyperscale CPUs were designed for renting cores, while AI-era CPUs like Vera must maximize token generation speed.

In this new world, no, AI doesn't rent cores. Yeah. AI generates tokens, and they sell tokens.

Huang says the new system is fully liquid cooled, cutting assembly time from two hours to five minutes, and is in full production.

Last generation, Grace Blackwell, was only 80% liquid cooled. Now, it's 100% liquid cooled. From two hours to assemble, it's now five minutes.

Huang expects NVIDIA to build and sell trillions of dollars of computing in coming years, requiring Taiwan's ecosystem to invest hundreds of billions.

If we're gonna go build and sell a few trillion dollars of computing over the next several years, this entire island has to be part of it.

Huang says the AI boom, plus AI and robotics in manufacturing, is the best opportunity to reindustrialize the United States.

During the, during the AI boom is our best chance to invest in the grid, invest in sustainable energy, invest in manufacturing again. There's so much demand.

Huang says the US lacks talent for world-class chip manufacturing and Taiwan is a strategic partner by sending engineers and capital.

It's not easy building up a world-class semiconductor manufacturing in United States because there's just not enough talent there. So they shipped a whole bunch of people, they're training people

Huang notes Chinese AI companies, though compute-starved, are forced to invent highly efficient architectures and remain well funded.

Yeah, but somehow, somehow, you know, while they're compute starved, it forces them to come up with these- - Yeah, exactly. - ... incredible architectures that are so efficient.

Huang argues export controls should be extremely selective, with US strategy optimizing a technology-and-wealth flywheel rather than denial.

So the question is, are you optimizing for someone else not having access to this technology, or are you optimizing for a flywheel? - Mm-hmm. - And I suggest that a more sophisticated long-term strategy is to optimize for a flywheel.

Huang says the American tech stack enjoys a global flywheel advantage partly due to English, which China lacks.

And so if you're an American company and you build on an American tech stack using American standards, you're fairly deployable throughout the world. - Mm-hmm. - That's not true with China. And so we have a global flywheel advantage. Technology without customers is useless.

Huang warns a platform flywheel built over decades can stop or slow very quickly, so the US must maximize platform reach.

Now, that flywheel is also really delicate.

Huang describes AI as a full technology stack where leadership in different layers, such as energy, sits in different countries.

In fact, AI is not a model. AI is a whole stack of technologies, and they're all contributing to AI. You can't have AI without energy.

Huang says the US and China should compete but do everything possible to avoid conflict, noting interdependence makes conflict harder.

I would say that we should do everything we can to avoid conflict. We should compete. That we can't do without.

Huang attributes Taiwan's central role in the global supply chain to its culture of generosity and hard work.

It's because of that fundamental generosity and the capacity to help other people. It's the reason why they're the supply chain for the world.

Huang compares 1990s foundries demanding IP licenses for capacity to early AI-company deals where capacity access required giving something away.

And kind of reminds me of, of some of the, the early deals with, with AI companies. You want to access the capacity, you have to license something away.

Huang says he chose TSMC over UMC for Morris Chang's values, viewing foundry selection as choosing a long-term partner rather than a product.

You really are choosing not what to buy, but who to work with. - Mm-hmm. - And so I was smart about making that decision.

Huang says NVIDIA's platform is large enough that its supply chain's stocks rose 300% last year while NVIDIA's stayed flat, and advocates leaving value for partners.

Nothing gives me more pleasure than last year when my entire supply chain's stock price went up 300%, mine stayed flat. Now, the fact that that happened is not something that I'm happy about, but it's because our platform is so large everybody could rise up to meet it.

Huang says NVIDIA plans a Taiwan building for 4,000 people and will hire about 3,000 more engineers there.

And, that building will hold 4,000 people. So we'll hire another 3,000 people.

Huang dismissed Hinton's 10-20% AI extinction risk estimate as unscientific and harmful.

That [figure] is not grounded on science, is not grounded on research. Just because it comes from a scientist doesn’t make it scientific. Those predictions are hurtful,

Huang argued AI alarmism does not serve the public good and called for evidence-based, scientific discussion of AI risk.

Don’t think for a second just because you’re an alarmist that you’re doing a social good. It is not true. So I think that we ought to just all be wiser, more mature, be evidence-based, be scientific.

Huang said all of Hinton's predictions about AI have been wrong.

I love [him]. I hate his predictions … All of his predictions have been wrong. Enough predictions,

Jensen Huang argues AI capability and safety must progress in tandem, praising the UK's scientific rigor in tackling AI safety as an engineering challenge.

“AI capability and safety must advance together,” said Jensen Huang, founder and CEO of NVIDIA. “The U.K. brings world-class scientific rigor to this essential engineering challenge.

Jensen Huang commits NVIDIA to providing $250 million in accelerated computing to UK researchers to build safeguards for increasingly powerful AI.

With this $250 million commitment, NVIDIA will put accelerated computing in the hands of U.K. researchers to develop, test and strengthen safeguards for increasingly powerful AI. This is how the U.K. turns research leadership into real-world solutions — and scientific progress into a future that benefits everyone.

Huang praised the White House accord and said Super Intelligence offers huge opportunity, with its future shaped by the confidence it earns.

Super Intelligence offers an extraordinary opportunity to advance discovery, productivity, security, health, and prosperity. The future of Super Intelligence will be shaped not only by what it can do, but by the confidence it earns.

Huang argues that achieving something extraordinary requires enduring suffering, since anything easy could be done by anyone.

“In order to do something incredible, you have to suffer. It’s not a walk in the park. If it’s a walk in the park, anybody could do it.”