AnswersPublished24 sources
China Could Meet 66% of Its Advanced-Chip Demand by 2035, Goldman Sachs Projects
Goldman Sachs projects China’s domestic supply of 7nm and below wafers could reach 410,000 per month by 2035—about 66% of projected demand of 619,000. The forecast assumes 46% annual supply growth, major semiconductor and AI investment, SMIC capacity additions and yields rising from 23% in 2026 to 75% in 2035.
Published byEdited with GPT-5.6 LunaImages generated with GPT Image 1.5
Research answer
Key takeaways
- Goldman Sachs projects China’s domestic supply of 7nm and below wafers could reach 410,000 per month by 2035—about 66% of projected demand of 619,000.
- The forecast assumes 46% annual supply growth, major semiconductor and AI investment, SMIC capacity additions and yields rising from 23% in 2026 to 75% in 2035.
- Lithography remains the central caveat: China’s reliance on DUV tools and lack of EUV access could keep its advanced chips more difficult and costly to produce, while CXMT’s DRAM progress may outpace its HBM capabilit...
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What does Goldman Sachs’s 2026 projection say about China’s ability to meet its domestic demand for advanced semiconductors—specifically, ho. Article summary: Goldman Sachs’s projection is not that China becomes fully self-sufficient in leading-edge chips by 2035. It is a high-growth but conditional scenario: domestic supply of 7nm-and-below wafers reaches about two-thirds of . Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
Goldman Sachs’s 2026 projection describes a sharp reduction—not the disappearance—of China’s dependence on overseas suppliers for advanced semiconductors. Domestic production of 7-nanometer-and-below wafers is projected to grow much faster than demand, reaching about 410,000 wafers per month by 2035 against projected demand of 619,000. That would cover roughly 66% of demand and leave a 34% supply gap. 2
6
The forecast is therefore best understood as a conditional industrial ramp. Its outcome depends on sustained investment, new SMIC capacity, substantially better manufacturing yields and continued access to the equipment and components needed to operate advanced fabs.
The headline numbers: faster supply growth, but a persistent gap
Goldman projects China’s domestic supply of 7nm-and-below wafers will grow at a compound annual rate of 46% between 2025 and 2035. Domestic demand is expected to grow at 17% annually over the same period, driven largely by AI-server workloads. 2
10
The resulting change in coverage is substantial:
- 2025: Domestic supply covers about 8% of demand, leaving a 92% shortfall.
- 2035: Domestic supply reaches about 410,000 wafers per month against demand of 619,000.
- Remaining gap: The deficit falls to approximately 209,000 wafers per month, or 34% of projected demand.
6
8
This arithmetic explains why the projection is significant without representing full self-sufficiency. China would produce most of the advanced-node wafers it needs under Goldman’s model, but it would still require imports or external production for more than one-third of demand.
AI investment supplies both the demand and the capital
AI infrastructure is important on both sides of the forecast. Goldman expects AI-server wafer demand to grow at about 42% annually, making advanced processors and related memory a major source of future semiconductor demand. 10
16
At the same time, Goldman raised its forecast for China’s semiconductor-industry capital expenditure to $82 billion by 2030, with double-digit annual growth expected before then. 17
20 The bank also estimates that Alibaba, Tencent, ByteDance and Baidu will together spend about $102 billion on AI-related capital expenditure in 2026.
20
22
That spending does not automatically translate into usable leading-edge chips. It can, however, support the broader build-out: fabs, manufacturing equipment, packaging, domestic suppliers and the data-center infrastructure that creates demand for AI accelerators. The strength of the forecast rests partly on whether this investment cycle continues long enough to fund repeated capacity expansions and process improvements.
SMIC is the forecast’s pivotal execution test
The advanced-logic scenario relies heavily on Semiconductor Manufacturing International Corporation, or SMIC. Goldman’s assumptions include adding roughly 30,000 to 50,000 advanced-node wafers per month each year from 2026 through 2031, followed by additions of about 20,000 wafers per month annually through 2035. 9
12
Capacity is only one part of the calculation. Goldman also assumes advanced-process yields improve from approximately 23% in 2026 to 50% in 2030 and 75% in 2035. 9
10
Yield measures the share of manufactured dies that meet specifications. A fab can have substantial installed capacity yet produce far fewer usable chips if defect rates are high. That makes yield learning a critical part of the projected supply increase, rather than a secondary operational detail.
The comparison with TSMC shows the cost challenge
Goldman’s assumed 75% SMIC yield in 2035 would still trail the more than 90% yield reported for TSMC’s mature 7nm process in the supplied reporting. TSMC has mass-produced 7nm chips since 2018. 13
The comparison matters because wafer volume alone does not determine competitiveness. Lower yields mean more wafer starts are needed to obtain the same number of good dies. For complex AI chips, that can raise costs, reduce effective capacity and make large-scale production more difficult even when nominal fab capacity is expanding.
Why lithography remains the binding constraint
The central limitation in Goldman’s outlook is lithography. China can produce 7nm-class chips using deep-ultraviolet, or DUV, systems and techniques such as multi-patterning. That approach is more complicated and generally slower and more costly than production using extreme-ultraviolet, or EUV, lithography. 2
3
China does not have access to ASML’s EUV tools under the export restrictions described in the reporting. As a result, additional funding and fab construction cannot by themselves close the technology, yield and cost gap. More equipment can expand nominal capacity, but the absence of the most advanced lithography tools increases the importance of process engineering, tool availability, component supply and yield improvement. 2
3
This is why Goldman’s projection points to narrowing dependence rather than complete independence. The 66% coverage figure assumes China can compensate for equipment constraints through scale, repeated process learning and continued access to DUV systems and related inputs.
CXMT could narrow the DRAM gap faster than the HBM gap
The forecast also includes a more favorable outlook for ChangXin Memory Technologies, or CXMT. Goldman projects that CXMT could supply approximately 50% of China’s DRAM demand by 2028, provided its planned capacity expansion, yield ramp and customer validation progress as expected. One reported capacity estimate rises from 270,000 wafers per month in 2026 to 447,000 in 2028 and 665,000 by 2030. 5
HBM is a harder test. Goldman-related reporting cites a forecast that CXMT could meet about 40% of China’s HBM demand by 2028, but other supplied reporting emphasizes that the company remains behind Samsung and SK hynix in HBM process technology, advanced stacking and packaging, reliability qualification and scale. 9
30
The distinction is important: progress in conventional DRAM capacity does not establish parity in AI-grade HBM. HBM requires not only memory-cell production but also demanding packaging, stacking and customer qualification. Restricted access to advanced lithography adds another challenge. CXMT’s likely path is therefore stronger in domestic DRAM substitution than in quickly matching the leading HBM suppliers.
What could derail the 2035 outcome?
Goldman’s 66% figure is a model scenario, not a demonstrated production result or a guaranteed industry target. It requires several assumptions to hold simultaneously:
- SMIC adds the projected advanced-node capacity.
- Yields rise from 23% to 75% over the forecast period.
- China sustains semiconductor and AI-related capital spending.
- DUV tools, components and supporting equipment remain available.
- Domestic customers continue adopting locally produced processors and memory.
- CXMT converts capacity expansion into qualified, commercially usable DRAM and HBM output.
3
5
9
A delay in capacity construction, weaker yields, equipment bottlenecks or slower customer qualification would leave a larger import requirement than the model assumes.
The bottom line
Goldman Sachs’s projection says China could make a major advance toward meeting its own advanced-chip needs by 2035. A 46% annual increase in 7nm-and-below wafer supply would lift domestic coverage from roughly 8% of demand in 2025 to about 66% in 2035, reducing the shortfall from 92% to 34%. 2
6
But the forecast does not describe full leading-edge self-sufficiency. Its most important variables—SMIC’s yields, access to lithography equipment and CXMT’s ability to qualify HBM—remain unresolved. China may become far less dependent on imported advanced chips while still facing a persistent technology and cost gap at the frontier.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "China Could Meet 66% of Its Advanced-Chip Demand by 2035, Goldman Sachs Projects"?
Goldman Sachs projects China’s domestic supply of 7nm and below wafers could reach 410,000 per month by 2035—about 66% of projected demand of 619,000.
What are the key points to validate first?
Goldman Sachs projects China’s domestic supply of 7nm and below wafers could reach 410,000 per month by 2035—about 66% of projected demand of 619,000. The forecast assumes 46% annual supply growth, major semiconductor and AI investment, SMIC capacity additions and yields rising from 23% in 2026 to 75% in 2035.
What should I do next in practice?
Lithography remains the central caveat: China’s reliance on DUV tools and lack of EUV access could keep its advanced chips more difficult and costly to produce, while CXMT’s DRAM progress may outpace its HBM capabilit...
Sources
scmp.comChina’s advanced chip deficit to shrink from 92% to 34% by 2035, Goldman says
techflowpost.com高盛研报解读：中国半导体数量自给率升至70%，2030年资本开支目标820亿美元