Nvidia CEO Jensen Huang promises a surprise next month, says: With a team like this ...

Nvidia CEO Jensen Huang has teased that the company is planning a surprise for its customers. In an interview with the Korean Economic Daily, he said that the US-based chip giant has “prepared several new chips the world has never seen before," and is planning to unveil "a chip that will surprise people" at Nvidia's GTC event in San Jose, California, which is set to be held next month.

The conversation took place after what Huang described as a “celebration of the world’s best memory [semiconductor] team,” following a dinner at a Korean fried chicken restaurant in California attended by engineers from SK Hynix and Nvidia. He also acknowledged that in the current environment “nothing is easy because all technologies are at their limits,” but added that “with a team like this, nothing is impossible.”

How SK Hynix is helping Nvidia with its next chip

Hunag noted that Nvidia and SK Hynix were “working very closely together” and operating as “one giant team,” adding that both sides had worked “really hard to meet the great challenge of Vera Rubin and HBM4.”

SK Hynix remains Nvidia’s primary supplier of high-bandwidth memory (HBM) chips, with HBM4 expected to be integrated into Nvidia’s Rubin architecture, which is scheduled to debut in the latter half of 2026.

In September 2025, SK Hynix said its sixth-generation high-bandwidth memory chips will deliver “the industry’s best data processing speed and power efficiency with the bandwidth doubled… and power efficiency improved by more than 40%.”

At that time, SK Hynix also stated that “the company expects to improve AI service performance by up to 69% when the product is applied.”

Earlier this year, SK Hynix warned that memory chip shortages were likely to continue into 2027, noting that consumer electronics could be more affected as manufacturing capacity is increasingly directed toward AI infrastructure projects.

The company has also announced plans to expand infrastructure investment to more than four times the level previously disclosed, with its M15X facility in South Korea expected to begin operations by mid-2027.

Last week, Nvidia made OpenAI's agentic coding tool Codex available to all 30,000 of its engineers, months after Huang told employees he wanted "every task that is possible to be automated with artificial intelligence to be automated."

The company-wide deployment announced by OpenAI was one of the largest enterprise rollouts of an AI coding assistant ever recorded. The ChatGPT maker said that it worked closely with Nvidia to deliver cloud-managed admin controls and US-only processing with fail-safes.

The latest version of Codex ran on the GPT-5.3-codex model. Dennis Hannusch, an engineer at Nvidia, said he started using the new model and called it "reaaally good."

"I keep expecting quality to drop deep into a session, but it doesn't," Hannusch wrote on X.

Another Nvidia engineer, Benjamin Klieger, said he was particularly impressed with the tool's context management and token efficiency, describing them as “probably the two most important advances for agents right now.