Nvidia: No Memory Oversupply Expected for Years Ahead, Open to CXMT Collaboration
nashnova research
CEO Jensen Huang said at GTC Berlin that Nvidia sees no memory oversupply for years and is open to working with China's CXMT. This means → Nvidia is signaling demand stays strong while keeping its supply-chain options deliberately wide.
Why does Nvidia say memory won't be oversupplied?
Huang's core argument: AI systems and other compute applications will keep consuming more memory as the market expands, absorbing new supply.
Nvidia executives highlighted two product categories — HBM (high-bandwidth memory, an ultra-fast memory designed for AI chips) and LPDDR (low-power double data rate, the energy-efficient memory used in phones and laptops) — both expected to see strong demand for years.
In plain terms = Nvidia believes AI eats memory faster than factories can make it, so "oversupply" is not on the horizon.
Why is the market worried about oversupply anyway?
South Korean media reported that Seoul expects a new wave of large-scale memory supply by 2029, potentially putting serious pressure on prices.
At the same time, China's CXMT (长鑫存储) recently made breakthroughs in DRAM technology, raising concerns that competition for SK Hynix, Samsung, and Micron will intensify.
This reflects the memory industry's recurring problem: breakthroughs expand capacity, and capacity colliding with a demand slowdown crashes prices. Both Korea and China ramping up at once is what makes this cycle feel different.
What did Nvidia actually say about CXMT?
Asked about potential collaboration with CXMT, Nvidia said it is open to the idea and noted it already does business with Chinese companies.
Nvidia added that China has significant infrastructure and technical capabilities, including manufacturing strength, that the semiconductor industry "must take seriously."
This means → Nvidia is not closing the door on Chinese memory suppliers — it is preserving strategic flexibility in its supply chain.
Did export controls actually help China?
Nvidia acknowledged that U.S. technology export restrictions have objectively accelerated China's chip and manufacturing development.
But the company also said American firms are prepared to compete.
In plain terms = the restrictions were meant to slow a rival down; instead they forced the rival to build its own capabilities. Nvidia concedes the point but insists it can still win.
Could CXMT actually enter Nvidia's supply chain?
"Open to working with" is a long way from an actual partnership — whether CXMT ultimately enters Nvidia's supply chain will be a key marker in the U.S.–China semiconductor contest.
Nvidia's stance reveals a balancing act: securing memory supply for AI products on one side, maintaining a contact window with Chinese suppliers on the other.
This signals that between geopolitics and commercial interest, Nvidia is choosing not to pick a permanent side — and that choice is itself a message.
市场有风险，内容仅供研究参考，不构成投资建议。