Technology · Article
Nvidia's Jensen Huang Dismisses Fears of a 2029 Memory Chip Glut
Published:
In brief
Nvidia CEO Jensen Huang told CNBC on October 22, 2026, that he sees no sign of a memory chip oversupply by 2029, pushing back on a theory that had helped trigger a sharp selloff in Samsung, SK Hynix and Micron shares. Speaking at Nvidia's GTC conference in Berlin, Huang said demand for high-bandwidth memory (HBM) would keep rising and that having more memory available was beneficial rather than risky.
Nvidia CEO Jensen Huang said on October 22, 2026, that he does not expect memory chip supply to outpace demand by 2029, directly contradicting a scenario that had rattled chip stocks for more than a week. Speaking to CNBC on the sidelines of Nvidia's GTC developer conference in Berlin, Huang said he saw no evidence supporting forecasts of a 2029 memory glut.
What Huang said in Berlin
Asked directly whether a memory oversupply could hit in 2029, Huang said he did not think so, according to the South Korean outlet SBS, which covered the Berlin remarks. Huang argued that demand for memory, especially high-bandwidth memory (HBM) used in AI accelerators, would keep expanding rather than leveling off.
The real beauty of memory is that the more you have, the better.
Hankyung, a South Korean business daily that also covered the Berlin comments, reported that Huang said HBM would continue to be used for years to come. SBS separately reported that Huang said Nvidia planned to keep buying large volumes of LPDDR, a lower-power memory type the company has deployed in data centers. News1 Korea's report carried a similar line, noting that Huang said more memory is better.
Why memory stocks had been falling
Huang's comments came after a sharp slide in memory-chip shares. Seoul-based outlet DigitalToday reported that, over five trading sessions, SK Hynix stock fell 26%, Samsung Electronics fell 22% and Micron declined about 11%; the report, published October 15 and updated October 23, 2026, did not specify the exact calendar dates of those five sessions. U.S. financial outlet 24/7 Wall St. reported a further move on October 20, when SK Hynix dropped 4%, SanDisk fell 3% and Micron eased 2% in a single session, a move it linked to a broader selloff on Seoul's Kospi index.
DigitalToday traced part of the pessimism to a newspaper column citing a Seoul National University professor and a Korea Semiconductor Industry Association executive, who argued that new memory supply could surge once a current investment supercycle matures past 2030, raising the risk of a later downturn. The same report noted that Elon Musk had separately said China would be able to produce memory chips at a large scale by 2028, adding to the unease.
| Company | Change | Period | Source |
|---|---|---|---|
| SK Hynix | -26% | Five sessions (DigitalToday report, Oct 15–23 window) | DigitalToday |
| Samsung Electronics | -22% | Five sessions (DigitalToday report, Oct 15–23 window) | DigitalToday |
| Micron Technology | -11% | Five sessions (DigitalToday report, Oct 15–23 window) | DigitalToday |
| SK Hynix | -4% | Oct 20 (single session) | 24/7 Wall St. |
| SanDisk | -3% | Oct 20 (single session) | 24/7 Wall St. |
The China factor: CXMT
A second source of pressure on the sector was Chinese manufacturer ChangXin Memory Technologies (CXMT), which announced plans for a server memory product built on new architecture, according to DigitalToday. The announcement revived concerns that CXMT could narrow the technology gap with Samsung, SK Hynix and Micron and take global market share.
Asked about CXMT in Berlin, Huang said Nvidia remained open to competing memory suppliers, according to SBS's report. Hankyung reported that Huang also praised China's broader semiconductor manufacturing capabilities and suggested he saw room for cooperation.
DigitalToday cited analysts who said the market may be overreacting: building new memory fabs typically takes three to five years, and announcing a product architecture does not guarantee the yields needed for mass production.
What is not yet known
None of the sources reviewed include an on-the-record response from CXMT, Samsung or SK Hynix to Huang's specific remarks, and no company has published updated 2029 supply forecasts since the Berlin comments. The original column that raised the 2030-plus supercycle argument, as summarized by DigitalToday, was an opinion piece rather than an official industry forecast.
Questions
- What exactly did Jensen Huang say about 2029?
- At Nvidia's GTC conference in Berlin on October 22, 2026, Huang told CNBC he does not expect memory chips to be oversupplied by 2029, saying demand for memory such as HBM will keep growing rather than leveling off.
- Why did memory chip stocks fall in October 2026?
- According to DigitalToday, SK Hynix, Samsung Electronics and Micron shares fell sharply over five trading sessions reported between October 15 and October 23, 2026, driven by a newspaper column warning of a possible 2030s memory supply surge and by concerns that China's CXMT could close the technology gap with established manufacturers.
- What is CXMT?
- ChangXin Memory Technologies (CXMT) is a Chinese memory chipmaker that announced a new server memory product architecture in October 2026, which DigitalToday reported as one trigger for the sector-wide selloff.
- Did Huang say anything about buying from CXMT?
- Per SBS's report of the Berlin interview, Huang said Nvidia remained open to competing memory suppliers, without committing to a specific purchase from CXMT.
- Is Huang's view shared by analysts?
- DigitalToday reported that some analysts consider the glut fears overdone, noting new memory fabs take three to five years to build and that announcing a product design does not guarantee mass-production yields.
How the world is telling it
- KoreanSouth Korean outlets (Hankyung, News1, DigitalToday) framed the story around the direct stakes for Samsung and SK Hynix, emphasizing stock-price damage, a local opinion column that fueled 2029/2030 oversupply fears, and Huang's remarks on China's CXMT as reassurance for domestic chipmakers.
Sources
- Hankyung — Jensen Huang on HBM demand and CXMT openness at GTC Berlin
- News1 Korea — "No Memory Oversupply in 2029... More Memory, the Better"
- SBS News (English) — Jensen Huang: No memory oversupply in 2029, open mind on CXMT
- 24/7 Wall St. — SK Hynix Sinks 4% as Seoul Selloff Revives AI Spending Fears
- DigitalToday — Why Samsung and SK Hynix shares tumbled: China chase and oversupply fears
This piece was prepared with artificial intelligence and passed an automated source check before publication. If you see an error, ask for a correction. Request a correction or reply · How we make this
Related
TechnologyArticle
Apple's iPhone Duo Foldable: Specs, Price and How It Stacks Up to Samsung
Apple's first foldable iPhone, the iPhone Duo, went on sale October 23, 2026, starting at $1,999, after preorder demand pushed shipping estimates into 2027 within minutes. The book-style device pairs a titanium hinge and IP68 water resistance — a first for this type of foldable — with a custom Samsung Display panel, intensifying competition with Samsung's existing Galaxy Z Fold line.
TechnologyNews
US Banks Say Unvetted AI Agents Are Logging Into Customer Accounts
Major U.S. banks including JPMorgan Chase, Bank of America and Charles Schwab say they are detecting third-party AI agents accessing customer accounts using credentials users handed over themselves, Axios reported on October 22, 2026. Banks say some agents have taken actions, such as moving funds, that appear disconnected from a specific user instruction, and are now blocking or restricting the practice.
TechnologyArticle
Why Big Tech Is Spending $730 Billion on AI Infrastructure in 2026
Amazon, Alphabet, Microsoft and Meta plan to spend roughly $730 billion combined on AI data centers and chips in 2026, up about 78% from 2025. The spree pushed Meta's free cash flow negative for the first time since its 2012 IPO and triggered a sharp tech-stock sell-off as investors question when, and whether, the spending will pay off.
TechnologyNews
PHP2Shell: Critical PHP Flaws Trigger Global Patch Rush and Site Shutdowns
Two critical PHP vulnerabilities nicknamed "PHP2Shell" let attackers run code on web servers without logging in, and PHP issued emergency patches on October 21, 2026. The flaws affect PHP installations behind roughly 70% of websites, prompting Japan's prime minister and cybersecurity office to urge companies to patch immediately or take sites offline, while Cloudflare and government agencies reported exploitation attempts within hours of disclosure.