Nvidia's Jensen Huang Sees No Memory Oversupply by 2029, Keeps Door Open to China's CXMT
On the sidelines of Nvidia's GTC Berlin conference, CEO Jensen Huang told CNBC that a memory supply surplus is unlikely by 2029, arguing that AI systems can always absorb more capacity, while signaling openness to Chinese supplier ChangXin Memory Technologies.
Published
Last updated / Fact-checked on
Huang spoke Thursday, Oct. 22, with CNBC's Arjun Kharpal. He said he does not expect an oversupply of memory by 2029. "One of the things that's really great about memory is that you can always use more of it," he said. He added that high-bandwidth memory (HBM), a form of RAM central to AI accelerators, "will be used for years to come."
He also highlighted Nvidia's early adoption of LPDDR — the low-power DRAM found in cellphones — inside data centers. "We're the first company in the world to put LPDDR memory … into the world's data centers, and it's enabled us to deliver performance that's twice as energy efficient as anything in the world today," Huang said. "We're going to continue to use abundant amounts of them as well."
Why the Comments Landed Now
Memory stocks had sold off over the prior week on fears that demand might weaken later in the decade. An Oct. 15 opinion column in a prominent South Korean publication urged the government to prepare for a slowdown in the AI boom. The column, citing two experts, warned of a "flood of new supply" that would turn the memory market into a "red ocean" in 2029, adding that "the supercycle could be over, and the industry could enter a downturn."
Elon Musk, CEO of SpaceX and Tesla, responded to that column by saying that China "will be making a LOT of RAM in 2028." Separate pressure came from China's ChangXin Memory Technologies (CXMT), which announced a new design architecture expected to improve efficiency. Investors worried the move could let CXMT take share from established suppliers SK Hynix, Samsung and Micron. A glut would typically lower prices and compress margins for those producers.
Openness to CXMT and Broader China Context
Asked whether Nvidia would partner with CXMT or use its memory, Huang replied, "We're open-minded." He noted, "We work with Chinese companies; just about the whole world works with Chinese companies in the supply chain somewhere." He added that "China is incredible at the fundamental infrastructure technology, and no country collectively is better at manufacturing than China, and so we have to take their capabilities very seriously."
Huang also addressed U.S. export restrictions on advanced semiconductors, saying they had ultimately accelerated China's own semiconductor manufacturing push. He added that American technology companies remain strong and ready to compete with anyone, according to Korean coverage of the interview.
Why It Matters for Computing and Chips
Memory, especially HBM and high-volume LPDDR, has become a binding constraint on AI accelerator deployment. Nvidia's systems consume large amounts of both, and sustained demand supports pricing and capacity expansion at the major suppliers. Huang's timeline — no expected surplus through 2029 — pushes back on near-term oversupply narratives that had pressured valuations. At the same time, his openness to CXMT underscores the global nature of the supply chain even amid U.S.-China technology restrictions. For system designers and cloud operators, the remarks reinforce that memory availability and efficiency gains will continue to shape what AI infrastructure can be built, and at what cost.
Key points from Huang's remarks
| Topic | Statement |
|---|---|
| 2029 oversupply | Unlikely; demand can absorb more memory |
| HBM outlook | Will be used for years to come |
| LPDDR in data centers | First mover; enables roughly 2x energy efficiency; continued heavy use |
| CXMT | "Open-minded" about partnering or using its products |
| China manufacturing | Exceptional at infrastructure technology and manufacturing; capabilities must be taken seriously |
| U.S. export rules | Accelerated China's domestic semiconductor push |
Implications
- Supports continued tight or balanced memory markets into the late 2020s amid AI-driven demand.
- Keeps Chinese suppliers as a potential option in Nvidia's broader supply considerations.
- Highlights energy-efficiency gains from LPDDR as one reason demand remains structural rather than purely cyclical.
- Reinforces that policy restrictions have spurred parallel Chinese capacity build-outs.
Sources
- CNBC, "Nvidia's Jensen Huang says he sees no sign of a 2029 memory glut: 'You can always use more'," Oct. 22, 2026: https://www.cnbc.com/2026/10/22/jensen-huang-nvidia-memory-selloff-shortage.html
- The Herald Business, "Jensen Huang shrugs off memory oversupply fears: 'More is always better'": https://biz.heraldcorp.com/article/10912035
- Yonhap Infomax, "Jensen Huang Says No Memory Oversupply Until 2029, Open to CXMT Products": https://en.infomaxai.com/news/articleView.html?idxno=144728