Jensen Huang Declares Nvidia an 'AI Systems Company': "Vera Rubin in Full Production... Supplied by Samsung and SK hynix" (Comprehensive)

Jensen Huang Delivers Nvidia GTC Keynote
Crowds Exceed 6,000 Attendees
"From Generative AI to the Age of Agent AI"
'Vera Rubin' Supercomputer Ushers in a New Era

Nvidia officially announced the commencement of mass production for its next-generation artificial intelligence (AI) accelerator, "Vera Rubin," emphasizing that the core of the AI industry is shifting from generative AI to "agent AI" (AI assistants). Originally a graphics processing unit (GPU) company, Nvidia declared its transformation into an "AI systems company," spanning not only AI semiconductors but also data centers, networks, CPUs, and software. The company also highlighted its collaboration with key Korean partners such as Samsung Electronics, SK hynix, Naver, and Hyundai Motor Group, as well as its global client partnerships.

Jensen Huang, CEO of Nvidia, stated in his keynote speech at Nvidia's annual AI conference "GTC Taipei 2026" held at the Taipei Music Center in Taiwan on the 1st (local time), "Vera Rubin has now entered full production," adding, "Vera Rubin is not a computer for AI, but a computer for agents."

Jensen Huang, CEO of Nvidia, is delivering a keynote speech at Nvidia's annual AI conference 'GTC Taipei 2026,' held at the Taipei Music Center in Taiwan on the 1st (local time). Photo by Jin-Young Kim.

CEO Huang predicted that the software architecture of the future will shift from a focus on applications to a focus on agents. He explained, "Agents observe, reason, plan, and use tools," emphasizing, "To handle agents, a new computing architecture is needed, one that integrates models, tools, runtime environments, and memory systems into a unified whole."

Nvidia explained that Vera Rubin was developed to address these changes. The company characterized Vera Rubin as an "AI supercomputer" with a multi-rack structure, integrating not only GPUs but also CPUs, networks, storage, and security processors, thus underscoring its expanded significance.

Declares Expansion Beyond GPU Company to Systems Company

On the 1st (local time), attendees are seated before Jensen Huang, CEO of Nvidia, delivers the keynote speech at Nvidia's annual AI conference 'GTC Taipei 2026' held at the Taipei Music Center in Taiwan. Photo by Jin-Young Kim.

The venue for the keynote was packed with more than 6,000 attendees. From as early as 8 a.m.—three hours before the event—people from all walks of life, including the general public, businesspeople, and journalists, lined up to enter. Even before the speech began, all 6,000 seats were already filled. Notable attendees included Madison Huang, Senior Director at Nvidia, and Chey Tae-won, Chairman of SK Group. As usual, CEO Huang appeared in his signature leather jacket and greeted the audience in Mandarin with "Dàjiā hǎo (Hello, everyone)" as he began his speech.

During his speech, he demonstrated that Nvidia’s business focus is shifting beyond semiconductor supply toward the broader infrastructure of AI data centers and AI factories. CEO Huang predicted, "In the future, every company will become an organization that operates agents," and added, "The next decade in the age of AI will be led by agents." He continued, "In the past, Nvidia was a GPU company, but now we have become a systems company. Customers are no longer buying computers; they are building AI factories. This is why Nvidia is once again beginning to reinvent itself."

He particularly emphasized the economics of AI infrastructure, repeatedly stating, "Compute is revenue." He explained that the core competitiveness of AI data centers lies not in the simple price of chips, but in throughput per watt (Token per Watt), power efficiency, and system reliability. CEO Huang stressed, "If you choose the wrong architecture, you cannot secure profitability even if chips are inexpensive," and added, "In the era of AI factories, computing performance directly translates into sales and profits."

On this day, Nvidia also unveiled the CPU "Vera," optimized for the age of agent AI. CEO Huang said, "Until now, CPUs have been designed for humans, but Vera is the first CPU designed for agents. In the future, the number of agents will far surpass that of humans."

Shout-outs to Korean Partners: Samsung, SK hynix, Hyundai Motor Group

NVIDIA CEO Jensen Huang is introducing collaboration with Naver Cloud during the 'NVIDIA GTC 2026' keynote speech held on the 1st at the Taipei Music Center in Taiwan. Screen capture from NVIDIA YouTube.

CEO Huang mentioned various global partners, including Korean companies, to showcase these relationships. He explained that SK hynix, Samsung Electronics, and Micron are supplying the sixth-generation High Bandwidth Memory (HBM4) used in Vera Rubin. He also added that it is built on TSMC’s 3-nanometer process and CoWoS-L packaging technology, as well as HBM4 memory.

HBM is a key memory that enables AI semiconductors to process data rapidly and is critical to AI server performance. Notably, Samsung Electronics recently became the first in the industry to pass Nvidia's HBM4 quality certification and began mass production and supply. This announcement serves as a reaffirmation that Samsung Electronics and SK hynix are vital partners supplying semiconductors for next-generation AI accelerators.

Additionally, while introducing global cloud partner cases, Nvidia featured Naver Cloud on screen alongside Google (USA), Yotta (Japan), Indosat (Indonesia), and GMI (Taiwan). Particularly, the presentation screen displayed the phrase "Nvidia♥Naver Cloud" with a red heart image. Among global cloud companies, the introduction of Naver Cloud with a heart symbol drew significant attention from the audience.

Hyundai Motor Group also appeared in the latter part of the keynote. In a robot performance video presented by CEO Huang, Genesis vehicles and IONIQ electric cars from Hyundai Motor Group were featured. As Nvidia increasingly focuses on robotics, autonomous driving, and physical AI, attention is also being paid to its cooperation with Hyundai Motor Group.

Meanwhile, Nvidia introduced its proprietary inference large language model (LLM), "Nemotron 3 Ultra," at the event. Nvidia plans to connect Nemotron with its AI software stack to help companies more easily build their own AI agents. Furthermore, Nvidia outlined its product roadmap for the next five years, including the next-generation accelerator "Feynman" scheduled for release in 2029, which will feature next-generation HBM. The company also announced plans to launch the "Rosa Feynman Spark" in 2030.

This content was produced with the assistance of AI translation services.