Nvidia CEO Jensen Huang told CNBC Thursday that a surplus of memory supply is unlikely by 2029, despite a recent sell-off in memory stocks driven by fears of oversupply and weakening demand. The comments come as investors react to warnings that the artificial intelligence boom may be slowing, potentially leading to a downturn in the memory market.
In an interview with CNBC's Arjun Kharpal on the sidelines of Nvidia's GTC Berlin conference, Huang argued that demand for memory will remain strong for years. He stated that high-bandwidth memory, a type of random access memory, will continue to be used extensively. Huang also highlighted Nvidia's use of LPDDR memory, a type typically found in cell phones, in data centers. He claimed this approach allows for performance that is twice as energy efficient as current alternatives, leading the company to use abundant amounts of the product.
Memory stocks have fallen sharply in the past week following an opinion column published on Oct. 15 in a prominent South Korean publication. The column urged the South Korean government to prepare for a deceleration in the AI boom. It cited two experts who predicted a flood of new supply would turn the memory market into a red ocean by 2029. They wrote that the supercycle could be over and the industry could enter a downturn by that time.
A glut in memory supply could lower prices and reduce profit margins for manufacturers such as SK Hynix, Samsung, and Micron. Elon Musk, CEO of SpaceX and Tesla, responded to the article at the time by stating that China will be making a lot of RAM in 2028.
Investor concerns were further fueled by an announcement from China's ChangXin Memory Technologies, or CXMT. The company revealed a new design architecture that could lead to greater efficiencies. Investors worry this development could allow CXMT to take market share from rivals SK Hynix, Samsung, and Micron.
When asked about partnering with CXMT, Huang said Nvidia is open-minded. He noted that many companies worldwide work with Chinese firms in their supply chains. Huang described China as incredible at fundamental infrastructure technology and stated that no other country is collectively better at manufacturing. He added that Nvidia must take China's capabilities very seriously.
Huang attributed China's advancements in the semiconductor industry to mounting U.S. regulations. He said these regulations pushed China to speed up its own chip technology and manufacturing. The U.S. currently restricts exports of its most advanced semiconductors to China, including top-end Nvidia chips. However, some Chinese firms are accessing this compute power through data centers in Southeast Asia, a loophole experts say has helped make Chinese AI models more capable.
Huang acknowledged that U.S. regulations have accelerated China's initiatives and the nationalization of semiconductor manufacturing capabilities. He stated there is no question that U.S. policies have sped up these developments. Despite this, Huang emphasized that American tech firms are mighty and prepared to take on competition from anybody.