[Exclusive] Jensen Huang: "AI investment is still in its early stages…there is no bubble"

Semiconductor INSIGHT

Exclusive interview with NVIDIA CEO

"We are at the outset of an infrastructure project worth tens of trillions of dollars"

Chimaek with SK engineers…"One team with Korea’s semiconductor industry"

"There is no artificial intelligence (AI) bubble. We are simply standing at the starting point of the largest infrastructure project in human history—one worth tens of trillions of dollars."

Jensen Huang, CEO of NVIDIA (pictured), said this in an exclusive interview with The Korea Economic Daily on the 14th. He met the reporter after having dinner that day with about 30 engineers from NVIDIA and SK hynix at 99 Chicken, a Korean-style fried chicken restaurant in Santa Clara, California. Despite the unscheduled interview request, he readily agreed, saying, "Ask whatever you like."

Huang described the evening gathering as "a get-together to celebrate the world’s best memory (semiconductor) team." Less than 10 days after a "chimaek meeting" at the same venue with SK Group Chairman Chey Tae-won and other executives on the 5th, he organized another occasion to encourage working-level staff from the two companies.

"SK hynix and NVIDIA are working very closely together," he said. "We have a great partnership." He also called the two companies "one gigantic team," underscoring that they share a common fate. The industry expects that the data transfer speed of HBM4 (sixth-generation high-bandwidth memory), which SK hynix is set to release soon, will be a key variable in resolving computational bottlenecks and maximizing performance for NVIDIA’s next-generation AI chip, "Vera Rubin." Huang said, "This team worked incredibly hard to take on the big challenges of Vera Rubin and HBM4, and they deserve a great evening enjoying soju and chicken," expressing strong confidence in the outcome of the collaboration.

Still, Huang signaled inevitable competition between SK hynix and Samsung Electronics over next-generation graphics processing unit (GPU) memory supply, saying, "Great companies that build great things will inevitably compete well." He added, "NVIDIA has competitors, and SK hynix also has competitors," and "It’s more worrying when there is no competition." The implication is that where there is no competition, there is no market.

"At next month’s GTC, we’ll unveil chips that will astonish the world"

Jensen Huang, CEO of NVIDIA, made it clear that the company will step up broad-based investment across AI infrastructure. "NVIDIA has great partners and amazing startups, and we’re investing across every stage (stack) of AI," he said. In December last year, NVIDIA brought in core technology and talent from Groq, a company that designs language processing units (LPUs) for inference, for $20 billion, and also invested $860 million in fusion startup Commonwealth Fusion Systems.

However, he avoided giving a direct answer to the question of whether he plans to make additional investments in AI foundation model companies such as OpenAI and Anthropic. Huang said indirectly, "AI is not just models; it’s an entire industry encompassing energy, semiconductors, data centers, the cloud, and the applications built on top." This is interpreted as meaning he intends to broaden the investment scope to the overall AI ecosystem rather than specific AI model companies.

He also expressed confidence about NVIDIA’s largest annual event, "GTC 2026," to be held in San Jose, California, from the 16th to the 19th next month. Huang said, "We’ve prepared several new chips the world has never seen before." While noting, "Nothing is easy because every technology is reaching its limits," he emphasized, "When teams like this (made up of NVIDIA and SK hynix memory engineers) come together, nothing is impossible."

Silicon Valley = Correspondent Kim In-yeop inside@hankyung.com