China Accelerates Lithography Equipment Development and AI Ecosystem Construction, But Faces Limitations

China is intensifying efforts to develop its own artificial intelligence models and achieve self-sufficiency in semiconductor equipment, presenting a new variable in the market competition landscape for the domestic semiconductor industry. Amidst U.S. export controls, the nation's push for indigenous lithography equipment and a complete semiconductor ecosystem could reshape the global semiconductor supply chain. However, analysts suggest that China faces significant hurdles in rapidly catching up to South Korea in the high-bandwidth memory (HBM) sector, a critical component for AI, which requires advanced process and packaging technologies.

“

China's technological capabilities are not only challenging the dominance of Dutch firm ASML in the global lithography equipment market but also emerging as a factor in the global AI supply chain.

According to various foreign media reports on July 29, China's state-owned enterprises are slated to deliver domestically developed immersion deep ultraviolet (DUV) lithography equipment to major Chinese semiconductor companies like SMIC, Hua Hong Semiconductor, and CXMT within this year. This development has led to analyses that China's technological capabilities are not only challenging the dominance of Dutch firm ASML in the global lithography equipment market but also emerging as a factor in the global AI supply chain. The DUV lithography equipment developed in China uses a 193-nanometer wavelength of light. To achieve the circuit density required for advanced processes, it necessitates multiple "multi-patterning" steps, which increases process complexity and reduces productivity and yield. Nevertheless, the industry views China's independent production of lithography equipment under U.S. export restrictions as a significant achievement.

“

A GPU that handles a thousand data points is worth little if the wire feeding it moves a hundred at a time, and that, he says, is exactly where Israeli engineering can win.

Beyond equipment localization, China is also rapidly building its AI ecosystem. Early this year, Chinese AI startup DeepSeek unveiled a "cost-effective AI model" designed to reduce reliance on HBM. Unlike U.S. hyperscale AI, which heavily utilizes NVIDIA's GPUs and HBM, China is constructing its AI ecosystem around its domestic companies' general-purpose memory. Chinese AI firms are also working to enhance compatibility with their domestic semiconductor ecosystem. For instance, AI startup Moonshot AI is optimizing its AI model "Kimi K(3)", considered a strong competitor to models from Anthropic and OpenAI, to run on Chinese-made AI chips.

“

South Korea appears to be 3 to 6 years ahead of China in most areas.

Despite these efforts in building a domestic semiconductor ecosystem, South Korea's dominance in high-value memory segments remains clear. HBM presents a higher barrier to entry than general-purpose memory, requiring not only fine-pitch processes based on EUV lithography but also advanced packaging and co-development with clients. This technological advantage underpins the partnerships between South Korean firms like Samsung Electronics and SK Hynix with U.S. tech giants for long-term memory supply contracts and AI data center collaborations. A report by the state-funded Korea Development Institute (KDI) School of Public Policy and Management titled "Innovation and Global Competitiveness of China's Semiconductor Industry" concluded that "South Korea appears to be 3 to 6 years ahead of China in most areas." An industry insider commented, "China and Samsung/SK Hynix are targeting different markets. The majority of memory customers are U.S. big tech companies, making it difficult for China to close the gap with Korean companies in the short term."

“

China and Samsung/SK Hynix are targeting different markets. The majority of memory customers are U.S. big tech companies, making it difficult for China to close the gap with Korean companies in the short term.