Where China's Chip Industry Stands
China consumes about 35% of the world's semiconductors but produces only about 15% domestically — and much of that domestic production uses foreign equipment and IP. The gap between what China buys and what it makes is the largest trade deficit in any single product category, exceeding $300 billion annually. Closing that gap has become a national priority with massive government funding behind it.
The challenge is staggering in scope. Semiconductor manufacturing requires equipment, materials, EDA software, IP cores, and process expertise that have been developed and refined over 50+ years by a concentrated set of companies in the US, Europe, Japan, South Korea, and Taiwan. Replicating that ecosystem domestically isn't a matter of throwing money at the problem — though China has certainly tried that approach, spending over $150 billion through the National IC Investment Fund and other channels since 2014.
SMIC: China's Foundry Champion
Semiconductor Manufacturing International Corporation is China's most advanced logic foundry. Its revenue is about $7 billion (versus TSMC's $75 billion), and its most advanced production node is officially 14nm FinFET. However, SMIC has demonstrated 7nm-class production — Huawei's Kirin 9000s processor, found in the Mate 60 Pro in 2023, was manufactured by SMIC on what appears to be a DUV-based 7nm-equivalent process.
That last point deserves attention. US export controls have denied China access to ASML's EUV lithography machines since 2019. Without EUV, manufacturing below 7nm using conventional lithography becomes impractical because of the multi-patterning complexity. SMIC's 7nm-class process reportedly uses quadruple patterning on DUV (193nm immersion) scanners — technically possible but extremely challenging in terms of yield and cost.
The yield is the real question. Industry analysts estimate SMIC's 7nm yield is significantly below TSMC's was at the same node, perhaps 50-60% for a mid-sized die compared to the 80-90% TSMC achieved. Lower yield means higher cost per good die and limited production volume. It's technically impressive but commercially constrained.
See also: GPU Architecture: CUDA Cores, Tensor Cores, and Ray Tracing .
SMIC's Expansion Plans
SMIC is aggressively expanding capacity at mature nodes where export controls are less restrictive. New fabs in Shanghai, Beijing, Tianjin, and Shenzhen are adding 28nm and above capacity. This makes strategic sense: the mature node market is large, the equipment is available, and Chinese demand for power management ICs, display drivers, MCUs, and analog chips can absorb the capacity.
For advanced nodes, SMIC's path forward is unclear. Without EUV and with tightening restrictions on DUV scanner sales (the Netherlands expanded controls in 2024 to cover some advanced DUV immersion tools), SMIC may struggle to advance beyond its current 7nm-class capability. The company has been stockpiling equipment — reportedly ordering as many DUV scanners as ASML would ship before restrictions tightened.
Hua Hong and the Second-Tier Foundries
Hua Hong Semiconductor is China's second-largest foundry, focused on specialty processes for power semiconductors, embedded flash, and analog/mixed-signal. Its technology is mature (130nm-55nm primarily) but addresses real market needs. Chinese EV makers, industrial equipment manufacturers, and consumer electronics companies need these chips, and domestic sourcing reduces supply chain risk.
Other Chinese foundries include Nexchip (formerly HeFei ChangXin Memory Technologies' logic division), CXMT-related facilities, and several smaller regional fabs that receive provincial government subsidies. The total capacity is significant at mature nodes, and China is on track to account for over 30% of global 28nm-and-above capacity by 2027.
Related reading: RISC-V Custom Extensions: Building Application-Specific Proc.
The EDA Tool Gap
Electronic design automation tools are a critical chokepoint. Synopsys and Cadence together control about 70% of the global EDA market, with Siemens EDA (formerly Mentor Graphics) holding most of the rest. All three are American or American-regulated companies subject to export controls.
Chinese EDA alternatives exist but are generations behind. Companies like Empyrean (now part of Primarius), X-Epic, and Semitronix offer tools for specific design steps, but no Chinese company offers a complete EDA flow capable of supporting advanced node design. You might find a usable schematic editor or a basic place-and-route tool, but the critical tools — timing analysis at advanced nodes, physical verification with FinFET-aware design rules, high-accuracy parasitic extraction — require billions of dollars in cumulative R&D that Chinese companies are still building.
The Chinese government has identified EDA as a priority, and investment in domestic EDA companies has increased substantially. But EDA tool development is as much about accumulated know-how and customer co-development as it's about engineering resources. Synopsys' tools have been refined through decades of collaboration with leading foundries, where tool development and process development happen in parallel. Replicating that ecosystem takes time regardless of funding levels.
Memory: CXMT and YMTC
ChangXin Memory Technologies (CXMT) is China's DRAM champion, producing DDR4 and LPDDR4X memory at its Hefei fab. Their technology is roughly equivalent to the 17-19nm node that Samsung and SK Hynix were at several years ago. It's behind the leading edge but functional and commercially shipped in some Chinese consumer electronics.
We covered a related topic in GAA vs FinFET: So Sánh Kiến Trúc Transistor Chi Tiết 2026.
Yangtze Memory Technologies (YMTC) made more headlines. Their 232-layer 3D NAND was competitive with Samsung and SK Hynix products when it was announced. Apple reportedly considered using YMTC NAND in some iPhone models. Then US export controls specifically targeted YMTC in October 2022, cutting off their access to American equipment and software. This has severely constrained their ability to expand capacity and advance to the next generation.
The memory situation illustrates both the progress China has made and the vulnerability of that progress to export controls. YMTC went from a credible competitor to an uncertain future within months of being sanctioned.
The Self-Sufficiency Timeline
China's goal of semiconductor self-sufficiency keeps getting pushed back. The original "Made in China 2025" target of 70% self-sufficiency in chips was quietly dropped when it became clear it was unreachable. Current realistic assessments suggest China might achieve 30-40% self-sufficiency in semiconductor content by 2030, concentrated in mature nodes and specific product categories.
Where China can realistically achieve self-sufficiency in the near term:
- Power management ICs and analog chips on mature processes
- Display driver ICs — companies like Chipone and Galaxycore already have significant market share
- CMOS image sensors — GalaxyCore and OmniVision's Chinese operations serve the domestic market
- MCUs and IoT chips on 28nm and above processes
- LED driver ICs and other commodity semiconductors
Where self-sufficiency remains distant: leading-edge logic (sub-7nm), advanced DRAM (sub-15nm), advanced NAND (200+ layers), high-end analog/RF, and the EDA tools and manufacturing equipment underlying all of them.
The honest assessment is that China can build a large and functional semiconductor industry at mature technology nodes, sufficient for many domestic applications. But for the most advanced chips that go into modern smartphones, data center accelerators, and AI training systems, dependence on the TSMC-ASML-EDA ecosystem will persist for the foreseeable future.