Speaking with CNBC at GTC Berlin, Jensen Huang pushed back on fears of a looming glut, saying a surplus by 2029 looks unlikely. His reasoning was straightforward: "one of the things that's really great about memory is that you can always use more of it," and HBM - a high bandwidth type of RAM - will remain in use for years.
He also pointed to Nvidia's approach with different memory types. Nvidia put LPDDR - the memory common in phones - into data centers, which he said enabled performance that is twice as energy efficient as anything available today. He added the company plans to keep using abundant amounts of LPDDR.
Memory names sold off over the past week after an Oct. 15 opinion column in a prominent South Korean publication cited two experts who expect a "flood of new supply" that turns the market into a "red ocean" in 2029. Their take: "The supercycle could be over, and the industry could enter a downturn," by 2029.
Additional pressure came from China's ChangXin Memory Technologies, or CXMT, which unveiled a new design architecture that could improve efficiencies. The update added to concern that CXMT could take share from SK Hynix, Samsung, and Micron. A glut would typically push prices down and squeeze margins at those producers. At the time, Elon Musk replied to the piece, saying China "will be making a LOT of RAM in 2028."
When keeping an eye on evolving markets, sign up for the free Market Briefs newsletter for five-minute daily explanations you can read fast
Asked about working with CXMT, Huang said he is "open-minded." He noted Nvidia works with Chinese companies and that global supply chains broadly include them. He also praised China's manufacturing edge, saying the country is exceptional at fundamental infrastructure technology and collectively unmatched in manufacturing, so its capabilities must be taken seriously.
Huang said U.S. rules have sped up China's chip push. Washington currently limits shipments of the most advanced semiconductors to China, including Nvidia's top-end chips. Some Chinese firms are accessing compute through data centers in Southeast Asia, a route experts say has been key to making Chinese AI models more capable. Even so, Huang called American tech firms "mighty" and ready to take on competition "from anybody."
Here is the tension for investors: memory is central to Nvidia's platforms, yet any future supply wave would weigh on SK Hynix, Samsung, and Micron margins. Huang's stance points to persistent demand - from HBM to LPDDR in data centers - while recent headlines have sharpened focus on potential competition from CXMT and the policy backdrop shaping China's ambitions. The bottom line: track memory supply signals, China's manufacturing advances, export rules, and any partnership chatter. Those are the levers moving valuations across the AI hardware stack.
If you want quick, clear context on the economy and markets, subscribe to the Market Briefs newsletter for a concise daily digest
