GYEONGJU, North Gyeongsang Province -- After announcing a landmark deal to supply 260,000 graphics processing units to South Korea, Nvidia CEO Jensen Huang faced the press in Gyeongju on Friday evening for a candid media session -- pausing between questions for sips of Coke, ice cream and a Korean chocolate snack to keep his energy up after a long day.
During the hour-and-a-half-long session Huang reiterated the importance of Samsung Electronics and SK hynix as “incredible partners,” that their complementary strengths are crucial to Nvidia's expanding AI infrastructure business.
He reaffirmed that the Rubin platform remains on track for mass production in the second half of 2026, hinting that the Korean chipmakers are likely to succeed in supplying their next-generation HBM4 chips for the platform.
Huang also expressed respect for Huawei as a formidable competitor while warning it would be “foolish to underestimate” China’s technological strength. Reflecting on Nvidia'’s soaring stock price, Huang emphasized that endurance and character are the company's true drivers of success.
Below are edited excerpts from Huang's media Q&A session.
Q. You said you will have a long-term relationship with Samsung and SK hynix, but the two differ in terms of technological capabilities, corporate culture and business models. Do you feel technological differences when you collaborate with them?
Jensen Huang: Samsung and SK hynix both have incredible technology capabilities. One company is much more focused, the other company is much more diverse -- and focus has its advantages, while diversity has its own advantages. We don’t need to choose between them, because we need both. We need the scale of both to support Nvidia’s business as well. You take them for granted, but I’ve been working with these companies for almost 30 years, and they are incredible partners -- my mightiest partners.
They will grow at an incredible scale to support Nvidia’s business next year, which will reach hundreds of billions of dollars. I’m grateful for their partnership, I’m grateful for their friendship, and I admire their capabilities. I need all their capabilities.
Q. President Trump said that Nvidia will be talking with China soon about the sales of computer chips. Although Blackwell was not made immediately available to China, could you comment on any specific products or businesses you would be discussing with China?
Huang: We’re always hoping to return to China, and I think that Nvidia in China is very good. It’s in the best interest of the United States; it’s in the best interest of China. I’m hopeful that both governments will arrive at a conclusion someday where Nvidia’s technology could be exported to China.
As an American company serving China, there are concerns about national security, that somehow the military could use American technology. But the fact is, China manufactures and produces plenty of its own GPUs. On one hand, I would say that Nvidia’s technology is a lot better. On the other hand, Chinese technologies are already plenty good for military applications. So I think the concern that American technology would be used by the PLA (People’s Liberation Army) for military use really doesn’t make sense, to be honest, because they have plenty of technology themselves.
It’s important for American tech stacks and American technology to participate in the Chinese market, where it is vibrant and dynamic with lots of developers.
Q. In September, Huawei unveiled its chip product roadmap, what was your reaction to that news last month? How far do you think Huawei is from Nvidia across all kinds of technology?
Huang: Nvidia is the world’s best AI infrastructure technology company. Our leadership in this field is extensive, it is deep. We have incredible scale. We move incredibly fast. We are inventive, we’re always inventing the future. We’re miles ahead.
However, it is foolish to underestimate the might of China and the incredible competitive spirit of Huawei. This is a company with extraordinary technology. They dominate the world’s 5G telecommunications standards and technology. They build amazing smartphones and chips, and they’re incredible at networking. So when they announced Cloud Matrix, I was not surprised that they were able to create such an amazing thing.
Q: Nvidia’s stock price has skyrocketed lately. Would you say it’s still a good time for investors to buy Nvidia shares now?
Huang: Nvidia is a great investment, and I heard from Jim Cramer: “Hold, don’t trade.” It’s probably more important to think about Nvidia from several perspectives. The first perspective: Why is this opportunity so unique and so large that it has taken all of humanity this long to invent the technology we have today? AI is the most complex technology the world has ever known -- and that’s why the opportunity is so large.
The second: What we do is incredibly hard -- that’s why Nvidia is the only company that does it.
The third: What about the quality and the character of the company? Is this a company that was founded a few years ago and became a flash in the pan due to accidental success? Or is it a company that has been galvanized through suffering, endurance, resilience, and dedication -- a company defined by its character? That character enables Nvidia to continue reinventing itself, to stay competitive, to keep creating the future.
One of the things I love about Korean culture is that your character is strong, because this is a culture that has suffered. And I love people who have suffered, because it is only through suffering that you achieve greatness.
Q. When will HBM4 qualification tests be concluded for Samsung Electronics and SK hynix, and is Rubin still on track for mass production in the second half of next year?
Huang: Rubin is on track for a second half production, nothing has changed. Rubin Silicon is back at Nvidia today and Rubin consists of six different very advanced chips. We have all the silicon in-house, we have all the systems in-house, and we’re preparing them for production. All of them are at Nvidia now and so I'm quite excited about delivering move in the second half of next year.
Q: You talked about combining AI with manufacturing in Korea. Which industrial area do you think will achieve physical AI the fastest, and what is the key condition for humans and robots to work together in factories?
Huang: Physical AI is AI that understands the laws of the physical world, for example, cause and effect, object permanence, inertia, momentum, et cetera. The industries that are related to the physical world are the largest industries in the world -- a hundred trillion dollars of industries will benefit from physical AI. Now, those industries also happen to be the industries of Korea -- making ships, making cars, making chips.
Korea will be able to grow because we could have a lot more AI labor. It will not replace humans, but will work with us, because there are many things that we can still do much better. I think the opportunity for Korea for physical AI and robotics is very significant, because Korea has the expertise, the technology as well as the market.
Q. To achieve physical AI and to establish an AI ecosystem, what kind of cooperation are you planning and what kind of results can we expect?
Huang: First thing we have to do is create AI, because without AI factories, you cannot produce AI. 260,000 GPUs, along with what’s already here in Korea, will make Korea one of the leading nations. We also need to have a rich suite of open models that encompass language, physical AI, robotics, autonomous vehicle systems, digital biology, chemistry -- a lot of different types of AI models that encompass all of the different industries.
We’ll work with startups and research universities like KAIST. One of the opportunities for Korea is that, now that you have this incredible AI infrastructure, you could attract companies from outside of Korea to operate their AI here.
Q. What is the most important factor to consider when making AI systems, and what are Nvidia’s core value regarding AI learning and trend?
Huang: Nvidia has to design six new chips every single year: GPU, CPU, two types of networking, and two types of scale-up and scale-out switches. The amount of software necessary to run this AI factory, this AI supercomputer, is unlike anything the world has ever seen. That’s why it’s so hard. That’s why Nvidia is such a large company, because the work we do is very serious, technically very deep.
We use a method called Extreme co-design. Co-design means we change the chips, the architecture, the systems, the software, the models -- all at the same time. When you have all of the technology in-house, you can change everything at once. We also optimize across everything -- from CPU to GPU, to networking, to switches, all the way across the entire data center. We are the only company in the world that does what I just described -- Extreme Co-Design at this large scale.
And we move really fast, every year we change everything, every year we invent new capabilities the world has never seen. That’s our core value, that we can design chips, software, AI, systems, and even AI factories -- all inside one company. It’s really quite rare.
herim@heraldcorp.com
