View: https://x.com/I_loves_deep_nn/status/1905770282243281130
For China's semiconductor industry, everywhere is full of news and speculation.
In my opinion, the big picture is settled
The Chinese semiconductor industry is grappling with a quintessential "chicken-and-egg" dilemma, particularly in its pursuit of advanced lithographic capabilities. Legacy ASML systems, such as older extreme ultraviolet (EUV) and deep ultraviolet (DUV) lithography machines, present an insurmountable reverse-engineering challenge due to proprietary failsafe mechanisms embedded within their architecture. Even Taiwan, with its unparalleled exposure to cutting-edge semiconductor fabrication through TSMC, has not succeeded in independently replicating these systems—underscoring the technological chasm. The intricacy of ASML’s photon projection systems, coupled with their reliance on highly specialized software and hardware integration, renders the notion of "cracking" these machines unfeasible for China in the near term.
For China to achieve self-sufficiency, two divergent paths emerge: either painstakingly iterate through generational advancements in photon-based lithography—mimicking the evolutionary trajectory of ASML and its peers—or pivot entirely toward a disruptive, paradigm-shifting approach to wafer patterning. The former is a multi-decade endeavor, constrained by both technical expertise and access to foundational intellectual property. The latter, however, offers a tantalizing prospect: with an estimated investment of $300–400 billion USD over a decade, China could channel its considerable pool of PhD-level research into novel lithographic techniques, potentially leapfrogging current standards and establishing a new technological benchmark. Recent advancements in light-source technologies, such as laser-produced plasma (LPP) for EUV, or emerging concepts like multi-beam electron lithography, could serve as viable starting points for such an initiative.
However, the Chinese ecosystem faces critical deficiencies in core competencies that undermine both strategies. Even with older ASML platforms, maintenance and operational proficiency remain elusive. These systems rely on intricate configuration software—often encrypted and remotely managed by ASML—leaving Chinese engineers incapable of performing deep-level servicing or modifications. Reports indicate that China has procured a significant number of these machines, yet lacks the domestic capacity to sustain their operational lifecycle independently. ASML’s decision to establish a maintenance hub in China, ostensibly to support academic and industrial research, highlights this dependency—a move that ensures continued oversight rather than technological transfer.
Optics, while a Chinese strength, is insufficient in isolation. The primary bottleneck lies in managing thermal distortions during the lithographic process—a domain where ASML’s expertise remains unrivaled and closely guarded. Heat-induced aberrations in wafer patterning, mitigated through advanced materials and real-time adaptive optics in ASML’s latest systems (e.g., the NXE:3400C), are a trade secret not even fully disclosed to strategic partners like the United States. Without breakthroughs in thermal management, software integration, and precision engineering, China’s ambitions risk stalling at the prototype phase. Propaganda touting self-reliance may sway domestic sentiment, but it holds no sway over the unforgiving physics of sub-7nm node fabrication.