Jensen Huang

Huang says Kazakhstan can move beyond exporting raw energy by converting it into intelligence and exporting intelligent solutions.

Instead of simply exporting energy, in the future you will be able to transform it into intelligence and export intelligent solutions,

Huang argues that broad participation from students, universities, scientists and industry in building the AI economy will generate demand.

You are inspiring young people, students, universities, scientists, and industry representatives to actively participate in the development of the artificial intelligence industry and AI economy. This is what will create demand,

Huang praises Kazakhstan's Kazakh LLM as a way to use national data to build its own AI.

I am very pleased to see that Kazakhstan already has its own large language model, Kazakh LLM, which enables the collection, processing, and use of national data to create its own artificial intelligence,

Huang says neocloud demand is off the charts and the only gap is working capital, which is scarce because AI demand is so broad.

The demand for the Neoclouds is off the charts. Their capabilities are extraordinary. And so what is missing in the middle is the working capital to fuel that growth, to fund that growth. And because there's just so much demand across so many sectors of AI, there's a scarcity of capital.

Huang rejects the idea of a 2029 memory oversupply, saying memory can always be used more and HBM and LPDDR will be needed for years.

I don't think so. Um, you know, one of the things that's really great about memory is that you could always use more of it and and um, of course, through architecture ideas and Nvidia owns the architecture of our systems. Through architecture ideas, we could shift around the resources uh, as necessary. But but uh HBM memory will be used for years to come and LPDDR.

Huang said he does not expect a memory glut by 2029 because AI can always absorb more memory.

One of the things that's really great about memory is that you can always use more of it

Huang said Nvidia's use of LPDDR in data centers doubles energy efficiency and it will keep consuming large amounts of memory.

We're the first company in the world to put LPDDR memory — what goes into your cell phones — into the world's data centers, and it's enabled us to deliver performance that's twice as energy efficient as anything in the world today, and so we're going to continue to use abundant amounts of them as well

Huang said China's manufacturing and infrastructure capabilities must be taken very seriously, and he is open-minded about working with CXMT.

China is incredible at the fundamental infrastructure technology, and no country collectively is better at manufacturing than China, and so we have to take their capabilities very seriously

Huang said US export regulations have unquestionably accelerated China's semiconductor self-sufficiency.

We've definitely accelerated China's initiatives, its nationalization of semiconductor manufacturing capabilities, and all of the industrial support that has been put into that sector, and so there's no question we've accelerated that

Huang says neo cloud demand is exceptionally strong, but a scarcity of working capital is holding back AI growth.

If you look at the fundamentals, the demand for the neo clouds is off the charts. What is missing in the middle is the working capital to fuel that growth … because there's just so much demand across so many sectors of AI, there's a scarcity of capital.

Huang compared controlling autonomous AI agents to looking after a baby in a playpen: contain it and monitor it.

"You contain the baby and you monitor it," Huang said Thursday morning on CNBC.

Huang expressed confidence that Nvidia's OpenShell would have handled past incidents of AI agents behaving unexpectedly.

"We're confident that it would have handled the incidents we've seen."

Huang argued that companies must act on AI safety immediately rather than wait for slow government regulation.

"Regulations take a while to come," he said.

Huang said the public can trust AI companies because commercial incentives require products to be safe.

"There's no such thing as a good product that's not safe," he said.

Huang acknowledged that new, unforeseen AI threats could emerge in the future.

"We don't know what we don't know about the future," he said.

Huang dismissed claims that AI will end human civilisation as untrue, calling such doomsday warnings made-up and cheap headline-grabbing.

When you say that it will be the end of the human civilisation, first of all, it’s not true,

Huang urged the world to approach AI with responsible optimism and keep building rather than stifling it, arguing it serves future generations and sustainability.

If you care about the future, past our own lives, if we care about the future of sustainability, this is the perfect time – build AI,” he said. “Don’t try to stifle AI.

Jensen Huang said last week that CoreWeave does not need a financial backstop from Nvidia.

While Nvidia CEO Jensen Huang said last week that CoreWeave doesn't need a backstop, some investors say the chip giant's further participation may be necessary if credit conditions continue to worsen.

Jensen Huang said demand for neocloud services is off the charts, though capital to fund the buildout is scarce.

demand for neocloud services is "off the charts."

Huang said Nvidia's revenue per gigawatt of AI capacity is now about $40 billion, up from roughly $18 billion a couple of years ago, because it sells networking, CPUs, storage and security alongside GPUs.

two gigawatts to us is like $80 billion

Jensen Huang argues ambitious companies will use AI-driven profit gains to enter new businesses and hire more people rather than cut staff.

真正有增长野心的公司，不会拿着 AI 当裁员省工资的工具。当 AI 把生产效率拉高、盈利变多之后，企业会把增量利润投入新赛道、开拓新业务，反而需要招聘更多人。

Jensen Huang said he is optimistic on CoreWeave and that its cost of capital is actually falling, despite the sell-off in its bonds.

optimistic, and their (CoreWeave's) cost of capital is actually decreasing.

Huang urged Europe to produce AI rather than merely consume it, saying it has the capacity to do so locally.

To be part of this AI economy, you don’t want to just be consuming the AI. You want to be producing the AI,” he told the 4,000-strong crowd at Nvidia’s GTC. “We can do all of that here in Europe.

Huang called humanoid robotics the next great industry and argued Europe's manufacturing heritage positions it to lead.

Humanoid robotics will be the “next great industry,” Huang said, arguing that Europe was well-placed to lead on it due to its manufacturing heritage.

Huang predicted that every company will combine closed models with customized open models.

Every company in the future will use both closed models as well as open models that you customize.

Huang said AI agents will expand software users and drive token consumption and compute demand higher.

He also predicted that agents would mean there will be more users of software tools than ever before, and said agents’ reasoning, iteration, and tool use are pushing token consumption and demand for computing—and Nvidia GPUs—higher.

Huang said the modular DSX Pod cuts idle compute time by coming online faster, calling it a complete transformation.

Huang said the DSX Pod would also slash the amount of time compute sits idle in a data center as it will come online quicker. “This is a complete transformation,” he said.

Jensen Huang says people now recognize that Nvidia's compute retains substantial residual value over time.

People are now recognizing Nvidia's compute has a great deal of residual value

Jensen Huang argues that AI safety work needs to be accelerated.

We have to accelerate AI safety

Huang says factory-integrated DSX Pods can cut the time from compute arrival to token production from four to five months to about one month.

By moving this into the supply chain, the speed by which we can turn on go from installation, arrival of computing, to production of tokens, will go from We think something along the lines of four to five months down to about a month.

Huang argues that pre-building systems in the supply chain will greatly reduce idle compute in datacenters and the revenue lost to delays.

By moving this into the supply chain the amount of idle computing, the amount of idle computing that's going to be sitting inside at your datacenters will reduce dramatically.

Huang says AI performance has improved a million-fold in ten years, far outpacing the roughly 100x Moore's law would have delivered.

In the course of that last 10 years, we've improved AI performance by 1 million times. If you would take a step back and think about what would Moore's law have done in 10 years' time, the first five years would have improved by a factor of 10. The second five years by another factor of 10, 100x improvement in 10 years' time.

Huang commits to making the Agile One SN humanoid reference robot available in Q1 next year, pre-assembled with compute and software.

This reference robot will be available in the first quarter of next year. And it will come pre-assembled with computer inside, software integrated, and it should be able to perform quite well right off the bat.

Huang says AI safety must be accelerated because safety and capability are inseparable.

We have to accelerate AI safety, and safety technology and capability to me is one and the same … As much as it is capable, you want it to be safe. The two of them are really one and the same.

Huang argues autonomous agents should be deployed under zero trust with minimal privileges, and that these practices should be adopted across the board immediately.

These agents are autonomous systems, and when you have an autonomous system on your computer … you should first start with zero trust, and give it minimum privileges, minimum access, minimum use of tools.

Huang claims Hugging Face's simulation showed NVIDIA's OpenShell would have prevented the attack on its site.

Hugging Face developed an exercise, a simulator, and they came to the conclusion: if OpenShell existed, their website wouldn't have been attacked that way.

Huang says NVIDIA is working with all labs and OS companies, and views the hacking incidents as unfortunate but transformative for the industry.

We're working with all the labs and working with all the operating system companies. I think the [hacking incidents were] unfortunate, but I think that it's very much transformative for the industry.

Huang argues AI factories create value only once they produce tokens, and that factory-built DSX POD units cut deployment from months to weeks so customers can scale and earn revenue sooner.

“An AI factory creates value when it starts producing tokens — not when construction is complete,” said Jensen Huang, founder and CEO of NVIDIA. “With NVIDIA DSX POD, we’re bringing factory-built precision and speed to AI infrastructure construction — reducing deployment time from months to weeks so customers can scale capacity and bring revenue-generating production online faster.”

Jensen Huang argues that AI capability and safety must advance together, and that security and control must be built into the computing architecture as enterprises deploy billions of agents.

“AI capability and safety must advance together,” said Jensen Huang, founder and CEO of NVIDIA. “As enterprises prepare to deploy billions of AI agents, safety and control must be built into the computing architecture. NVIDIA created OpenShell to give agents the freedom to do useful work within clearly defined boundaries — contained, monitored and under control.

Huang says telecom networks will evolve from carrying conversations to carrying intelligence, with operators turning land, power and connectivity into distributed AI infrastructure.

Telecom networks were built to carry the world’s conversations. Now, they will also carry the world’s intelligence. Leading global operators are turning their land, power and connectivity into a distributed AI infrastructure — bringing intelligence closer to the industries, people and robots that use it.

Huang says NVIDIA and Nokia are extending accelerated computing into the radio network itself, laying the foundation for AI-native 5G.

With Nokia, we’re extending accelerated computing into the radio network itself, creating the foundation for AI-native 5G.

Huang says robotaxis are the first mass-market application of robotics and physical AI, and that Hyperion with Halos offers a full-stack platform toward a future where everything that moves is autonomous.

Robotaxis are the first mass-market application of robotics and physical AI, reinventing how people and goods move. NVIDIA Hyperion, together with our Halos safety operating system, gives our partners a full-stack robotaxi platform designed to achieve full autonomy and the highest levels of safety, bringing us closer to a future where everything that moves is autonomous.

Huang says general-purpose robotics is advancing extremely fast and that robotics' ChatGPT moment is fast approaching.

General-purpose robotics is advancing at extraordinary speed — the ChatGPT moment for robotics is fast approaching,

Huang says the open full-stack Isaac platform gives developers models, data, simulation and compute to bring physical AI to the largest industries.

With NVIDIA Isaac, our full-stack open robotics platform, we are giving developers the models, data, simulation tools and computing systems to bring physical AI to the world’s largest industries.

Jensen Huang says AI is driving the largest infrastructure buildout in human history, with demand extraordinary and supply capacity as the binding constraint.

We are in the midst of the largest infrastructure buildout in human history. Demand is extraordinary. Capacity is the constraint,

Huang argues Europe should produce AI rather than merely consume it, and says it has already begun doing so.

To be part of this AI economy, you don’t want to just be consuming the AI; you want to be producing the AI,” Huang said. “We could do all of that here in Europe, and in fact, we’ve already started.

Huang says the modular DSX POD will transform the supply chain and speed up data center construction.

This is a complete transformation of the world supply chain,” Huang said. “This will change the way that people install and build data centers, and hopefully, as a result, velocity goes up.

Huang stresses that agent security is paramount, with minimal privileges and monitoring.

The most important thing is to make sure the agent is secure,” Huang said. The autonomous agent can access files, resources, networks and tools, so its privileges should be minimal and it should be monitored.

Huang says companies can post-train open models like Nemotron on their own data to reach super intelligence with today's technology.

You could start with an open model, something that’s incredibly good like Nemotron, and with your own data, post-train it for your use case and achieve, of course, super intelligence,” Huang said. “That is a very sensible thing to do, and all the technology exists today.

Huang predicts AI will revolutionize telecommunications and greatly improve spectrum efficiency.

Now CUDA makes phone calls,” Huang said. “AI will revolutionize the way we do telecommunications. It will enhance our spectrum efficiency tremendously.