Jensen Huang calls on DRAM manufacturers to go ahead and expand production: "No matter how much you make, I'll buy it all."

```

Nvidia CEO Jensen Huang has sent a strong signal to global memory chip manufacturers: no matter how much production expands, Nvidia will take it all.

On Monday, Huang stated at the Morgan Stanley Technology Conference that supply shortages for chips are “excellent news” for Nvidia, because limited resources prompt customers to directly opt for the highest-performing solutions. He spoke directly to DRAM manufacturers: “However much capacity you expand, we (Nvidia) will use it all.”

This statement is seen by industry insiders as a strong endorsement of the DRAM industry, as well as an indication that demand for storage resources amid the AI compute arms race will continue to climb.

The industry believes that behind Huang’s statement, is Nvidia’s upcoming new generation Vera Rubin platform’s huge demand for storage resources. As Nvidia fully pushes the new platform, contract manufacturing partners like Foxconn, Quanta, and Wistron are also expected to see improved order prospects.

Supply constraints become an advantage, Nvidia goes all-in on procurement

Huang explained his unique logic on resource shortages: In a scenario where resources like land, electricity, and construction for data centers are all limited, customers become more cautious in their purchasing decisions and are more likely to choose the strongest solution in one go. This trend is especially pronounced for memory.

His comments mean that even if DRAM prices continue to rise due to supply tightness, Nvidia will maintain strong procurement efforts and will not cut inventory due to cost pressure. This gives Samsung, SK Hynix, Micron—the three global DRAM giants—clear visibility on demand.

Nvidia’s thirst for storage resources is closely tied to its new-generation AI platform technology route. For example, the existing GB300 supports high-bandwidth memory (HBM) up to 288GB, a significant increase over the GB200’s 192GB.

Entering the Vera Rubin era, HBM capacity remains at 288GB, but specifications will upgrade from HBM3E to HBM4. HBM4 uses a 16-layer stack design, which is more complex than HBM3E’s 12-layer stack, naturally leading to higher yield losses and greater consumption of storage resources per unit output. This is expected to further exacerbate the DRAM supply shortage.

Three major DRAM manufacturers expand production, consumer-end gaps difficult to fill in the short term

Although Samsung, SK Hynix, and Micron have all initiated expansion plans, industry insiders note that the three manufacturers are currently focusing primarily on HBM and enterprise-grade DRAM, and the supply shortage for consumer DRAM will likely remain unresolved for at least the next year.

Meanwhile, prices for HBM as well as data center-use LPDDR4, LPDDR5, and the new DDR6 are expected to continue rising. The persistently tight supply-demand structure supports the profitability outlook for memory chip manufacturers.

Analysts are optimistic that Nvidia’s continued expansion of storage procurement and push for new-generation AI platform opportunities will drive momentum throughout its supply chain. Foxconn, Quanta, and Wistron, core assembly partners for Nvidia’s AI servers, are expected to benefit from this round of demand expansion.

Risk Warning and DisclaimerThe market involves risks; investment needs caution. This article does not constitute personal investment advice and does not take into account individual users’ specific investment goals, financial positions, or needs. Users should consider whether any opinions, views, or conclusions contained herein fit their unique circumstances. Invest accordingly, and you are responsible for your own actions. ```