Key Takeaways
- Chinese Academy hit 3.42% EUV efficiency in 2025, narrowing ASML’s chipmaking lead.
- ASML’s 2019 China ban still limits sub-7nm chips, pushing domestic EUV efforts toward 2028.
- Alibaba unveiled 2.4T-parameter Qwen3.8 as T-Head expands post-CUDA AI tools in 2026.
The most valuable bottleneck in advanced chipmaking is the EUV light source, and for years it has sat behind a single Dutch gatekeeper: ASML. Now a Chinese Academy of Sciences team at the Shanghai Institute of Optics and Fine Mechanics, led by ex-ASML scientist Lin Nan, has published results for a laser-produced plasma EUV platform that hits internationally competitive parameters, including a reported 3.42% conversion efficiency in a March 2025 paper. The Diplomat notes the power output is still only about 100 to 150 watts, well short of ASML’s commercial benchmarks, but the direction matters for a country blocked from buying ASML’s most advanced EUV tools since 2019. In Shanghai this week, Alibaba is also leaning into the same thesis from the software side, previewing its 2.4 trillion-parameter Qwen3.8 and having T-Head pledge to open-source SAIL as a CUDA alternative.
Cracking the EUV bottleneck
For US readers tracking semiconductors, the real drama is not just new chips, it is the tools that make them. A year after a key report surfaced, China is still trying to loosen a choke point that has shaped everything from iPhone supply chains to Nvidia-powered AI clusters. The latest progress centers on the light source needed for extreme ultraviolet lithography.
A team at the Chinese Academy of Sciences’ Shanghai Institute of Optics and Fine Mechanics, led by Lin Nan, built an EUV light-source platform that a research paper describes as operating at internationally competitive parameters, a step the May 12, 2025 report framed as a barrier giving way. Lin previously worked at ASML in the Netherlands, including as a competence owner for light sources used in metrology, a detail that helps explain why US officials watch these technical papers so closely.
A strong lab result, short of factory reality
In March 2025, Lin’s group reported an LPP (laser-produced plasma) EUV setup driven by a solid-state 1-micron laser, rather than the CO2 lasers used in commercial systems. The paper put its peak conversion efficiency at 3.42%, beating earlier academic results from ETH Zurich (1.8% in 2021) and a Dutch research center (3.2% in 2019). Commercial EUV light sources are still around 5.5% efficient.
Power output remains the part that separates prototypes from production lines. A recent analysis said the EUV light generated by China’s methods has been around 100 to 150 watts, versus 600 watts for ASML’s current commercial benchmark. That gap matters because higher power enables higher wafer throughput, which is how fabs make money.
Export controls, reverse-engineering, and the 2028 target
US export pressure has kept China from buying the most advanced ASML EUV systems since 2019, and EUV is a key requirement for chip nodes below 7 nm. One report put China’s lithography-machine market as 99% controlled by ASML and Japan’s Nikon and Canon, underscoring how much of the supply chain sits outside Beijing’s reach.
Reuters reported in December 2025 that researchers in Shenzhen had assembled an operational prototype EUV lithography machine in early 2025 by reverse-engineering ASML technology, using parts salvaged from older tools and other export-restricted components. The machine was not producing functional chips, but the ambition was blunt: “The aim is for China to eventually be able to make advanced chips on machines that are entirely China-made.”
Alibaba’s parallel push: bigger models, tighter stacks
At the World AI Conference in Shanghai, Alibaba previewed Qwen3.8-Max-Preview on July 19, 2026, describing it as a multimodal model with 2.4 trillion total parameters using a sparse Mixture-of-Experts design, per the July 20, 2026 write-up. Alibaba has not yet disclosed benchmark results, license terms, or how many parameters are active at inference time.
A day earlier, Alibaba’s chip unit T-Head said it would open-source SAIL, the software stack behind its Zhenwu chips, positioning it as an on-ramp away from Nvidia’s CUDA ecosystem. That announcement follows T-Head’s May 20, 2026 unveiling of the Zhenwu M890, which it said delivers 0.6 PFLOPs FP16 with 144 GB of HBM3, and had already shipped 560,000 units to more than 400 corporate customers by April 2026.
