Nvidia, SK Hynix deepen partnership for next-generation AI infrastructure

Nvidia CEO Jensen Huang and SK Group Chairman Chey Tae-won met Sunday on the sidelines of GTC Taipei 2026, marking the occasion together as SK Hynix recently crossed the $1 trillion market capitalization threshold.

Chey also attended Huang's GTC keynote address in person, taking stock of the rapidly shifting frontiers of AI technology. The Taiwan meetings are seen as reinforcing both companies' sense of each other as partners in building out the AI architecture of the future.

SK Hynix shared photos of the meeting on social media Monday. Senior executives from both sides were present — SK Group was represented by SK Hynix CEO Kwak Noh-jung and SK Telecom Chief Technology Officer Jeong Seok-geun, among others.

Huang and Chey celebrated SK Hynix's achievement of a $1 trillion market capitalization, which the company reached on May 27.

"The leaders of both companies came together to share the significance of that milestone," SK Hynix said in its post, calling the gathering "a meaningful occasion to reflect on what we have achieved together in AI memory and to reaffirm our commitment to opening a new horizon in AI infrastructure."

Huang also referenced SK Hynix's share price surge at a Korea Partner Night event held during GTC Taipei.

At a press briefing for South Korean media, Huang said SK Hynix had recently become a $1 trillion company. "I am very happy to see their success," he said.

Huang also highlighted what sets SK Hynix apart as competition intensifies over sixth-generation high-bandwidth memory, or HBM4, supply.

"For HBM supply, performance, quality, reliability and supply capability all matter," Huang said. "That is why we need to work closely with SK."

SK Hynix holds a 50 to 60 percent share of the global HBM market, having secured a dominant position in Nvidia's HBM supply chain. The company exclusively supplied its fourth-generation HBM3 chips for Nvidia's H100 GPU after beginning mass production in 2022, then pulled further ahead of rivals in 2024 by becoming the first in the world to mass-produce a 12-layer stacked HBM3E. SK Hynix completed development of HBM4 in September last year and has since established a mass-production framework.

The Taiwan meeting is seen as a sign that the SK Hynix-Nvidia partnership is evolving into a long-term collaboration aimed at jointly shaping the AI infrastructure era.

Meanwhile, Chey followed Huang's keynote closely throughout, using the presentation to assess firsthand what role SK Hynix should play as the AI ecosystem undergoes rapid transformation.

During the keynote, Huang outlined the evolution of GPU-based accelerated computing and the latest trends in AI innovation. He also unveiled the mass-production roadmap for Vera Rubin, Nvidia's next-generation AI platform, and detailed the company's collaboration with Asia-Pacific partners.

Huang officially announced that Vera Rubin "has entered full production," adding that the platform incorporates TSMC's 3-nanometer process, advanced packaging, and HBM4 from Micron, SK Hynix and Samsung Electronics.

He also presented business results in physical AI — including self-driving cars and industrial robots — along with strategies for building out AI factories and an open-source AI model ecosystem. SK Hynix was cited as an example of an AI factory.

The two leaders have been meeting with increasing frequency. They convened in Silicon Valley in February to discuss AI infrastructure cooperation, then met again at GTC in San Jose in March to flesh out the scope of their collaboration.

An SK Hynix official said the company also plans to introduce its "Full-Stack AI Memory Creator" vision to major global customers during the trip.

By Lee Jeong-wan and Park Ji-young (Taipei)