Test report DSG-4208 · Rev E · tested November 17, 2026
Supply Chain & PolicyDevice under test
Hanwha sees GPU shortage lasting through Nvidia’s fiscal 2028
Hanwha says AI accelerator and memory demand should stay firm despite tighter U.S. China controls, which could constrain Chinese DRAM and NAND capacity and support pricing.
- Read
- 3 min
- Words
- 681
- Node
- 7nm
- Operator
- Grace Kim
Spec summary
- Nvidia projected its GPU shortage would continue through fiscal 2028.
- Nvidia’s related cloud-capacity purchase commitments stood at $36 billion in fiscal Q2 2027.
- Traditional hyperscalers generated 54.7% of Nvidia data-center revenue; other customer groups generated 45.3%.
- Nvidia’s planned AI infrastructure financing platform targets more than $500 billion in third-party capital.
- Nvidia guided to about 70% revenue growth in fiscal 2028, below customer demand forecasts.
By Grace Kim
Nvidia expects the shortage of its latest graphics processing units (GPUs) to continue through the end of fiscal 2028, according to an August earnings release cited by Hanwha Investment & Securities (003530.KS).
Hanwha expects demand for AI accelerators and memory chips to remain firm despite the departure of Nvidia Chief Executive Jensen Huang and tighter U.S. restrictions on semiconductor exports to China. Demand for Nvidia’s latest GPUs far exceeds supply even if China is excluded.
Financing changes, not the demand cycle
Hanwha said on the 17th that reduced vendor financing, a market concern since Nvidia’s management change, is unlikely to spread into a broader contraction in AI investment.
The brokerage treats Nvidia’s equity investments and cloud-capacity purchase commitments to AI cloud operators as a company-wide strategy to expand the GPU ecosystem, not a course set by any individual chief executive. Related commitments stood at $36 billion in Nvidia’s second quarter of fiscal 2027.
Chinese restrictions could constrain memory output and improve the supply-demand balance for Samsung Electronics (005930.KS) and SK hynix (000660.KS). “In that it constrains supply growth from Chinese memory makers, it could work positively for the supply-demand balance in Korea’s memory industry,” Park Jun-young, an analyst at Hanwha Investment & Securities, said.
Traditional hyperscalers, meaning large cloud-computing operators, generate 54.7% of Nvidia’s data-center revenue. AI cloud providers, model developers, general enterprises and sovereign AI projects contribute the remaining 45.3%.
Hanwha believes reduced financing for some newer AI cloud operators, known as neoclouds, would not erase that broader demand. Other customers seeking to secure GPUs could replace the lost volume, the brokerage said.
Nvidia is also building an AI infrastructure financing platform with outside financial institutions. It aims to mobilize more than $500 billion in third-party capital over an extended period.
Supply, not demand, sets the near-term ceiling
The bottleneck lies closer to supply. Nvidia projected in August that the shortage would continue through fiscal 2028 and had guided to revenue growth of about 70% for that year. Customer demand forecasts far exceeded the 70% figure.
Because current capacity cannot satisfy all customer demand, Hanwha sees considerable room to reallocate GPU volumes blocked by China-related restrictions to unmet buyers in other regions. Export controls could change Nvidia’s revenue mix by customer or the timing of investments without structurally reducing AI accelerator output.
U.S. bills target different links in the chain
The Chip Security Act and AI OVERWATCH Act, both under consideration in the U.S. Congress, focus on blocking China’s access to advanced AI chips and preventing circumvention through third countries. They address access to chips rather than Chinese production capacity.
The MATCH Act would tighten controls on manufacturing equipment, components and maintenance services for Chinese chipmakers, constraining their production capacity itself. If it takes effect, it could slow DRAM (dynamic random access memory) and NAND flash expansion at ChangXin Memory Technologies and Yangtze Memory Technologies.
The measure could restrict new purchases of lithography (patterning), etching (material removal) and deposition (layer formation) equipment for advanced processes. It could also limit maintenance of existing equipment and process conversions between manufacturing recipes, slowing both added capacity and production upgrades.
Rising memory demand from AI infrastructure investment could turn that constraint into support for global DRAM and NAND prices. Slower Chinese supply growth would improve the balance for the Korean producers.
Hanwha maintained its medium- to long-term demand outlook for high-bandwidth memory (HBM) and conventional memory, as well as its “positive” view on the Korean semiconductor sector. Chinese restrictions may alter Nvidia’s revenue mix or investment timing, but they should not structurally damage memory demand.
What does this change for memory buyers?
Tighter U.S. controls could slow Chinese DRAM and NAND capacity growth just as AI infrastructure demand rises, supporting global memory prices and the supply-demand balance for Samsung Electronics and SK hynix. GPU demand would remain strong enough, in Hanwha’s view, to keep the broader shortage running through fiscal 2028.
via en.sedaily.com (Source)
More from Grace Kim
Show full bio
Market editor covering regulation, policy and their impact on hardware layer of artificial intelligence at Die Signal.
665 articles
Same lot · LOT-362D
- DSG-820310nmNvidia reserves AI accelerator capacity through 2028
- DSG-44167nmNvidia’s $500 Billion AI Plan Raises Korea and Export-Control Risks
- DSG-93143nmAmazon Weighs Sale of $8 Billion in Nvidia AI Chips
- DSG-24585nmChinese AI firms bought 1 million Nvidia GPUs in one year
- DSG-381914nmAmazon Seeks to Offload $8bn of Nvidia Chips to Investors