By Yang Jie and Sherry Qin

Nvidia's Jensen Huang says Marvell Technology could be the next chip firm to join the trillion-dollar club, predicting a surge in demand for artificial-intelligence hardware fueled by the emergence of autonomous models.

Speaking at a trade show in Taipei alongside Marvell CEO Matt Murphy, Huang mapped out the future of AI infrastructure, highlighting a strategic transition from copper cables to optical communications.

"Useful AI has arrived," Huang said, citing the rise of autonomous AI agents that can run workflows and solve problems as the reason why demand for Marvell and Nvidia's respective products is "going through the roof."

The high-profile joint appearance at Computex in Taiwan on Tuesday underscored the critical role of next-generation network fabrics in the AI era.

Murphy, whose Santa Clara, Calif.-based company specializes in data infrastructure semiconductors and high-speed networking technology for data centers, opened his keynote by asserting that the next major wave of AI innovation will be driven by interconnection.

Focusing solely on processor or memory fails to capture the full picture of hardware efficiency, said Murphy.

Addressing the technical challenges of scaling massive AI data centers, Huang said that computing is becoming increasingly disaggregated and distributed, and to tie these sprawling systems together, the industry must rely heavily on advanced connectivity.

"That's the reason why Marvell is so essential," the Nvidia CEO said, telling Murphy: "That's why you're going to be the next trillion-dollar company."

Nvidia announced a strategic partnership with Marvell in March, saying that it has invested $2 billion in the company.

Marvell's stock, which finished 7% higher on Monday, surged in off-hours trading on the Blue Ocean alternative system, climbing over 15%.

Speaking on the industry's shift from copper wiring to silicon photonics for data transmission, Huang advocated for a pragmatic, cost-effective way to maximize the lifespan of existing copper architecture within AI systems while deploying more expensive optical links only where technically essential.

As AI workloads become more complex, copper is hitting its physical limits, gradually pushing companies to turn more to optoelectronics.

"You use optics wherever you must, [and] you use copper wherever you can," said Huang.

Write to Yang Jie at jie.yang@wsj.com and Sherry Qin at sherry.qin@wsj.com

(END) Dow Jones Newswires

June 02, 2026 02:23 ET (06:23 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.