Siemens and NVIDIA Speed Up Computational Lithography for Advanced Chips
Siemens and NVIDIA are combining NVIDIA cuLitho with Siemens Calibre software to accelerate computational lithography workflows. The collaboration provides at least 10x acceleration for key lithography simulation workloads, with the headline result reaching up to 12x. The companies say this can help enable advanced patterning techniques such as curvilinear inverse lithography for next-generation AI semiconductors.
Siemens is pairing its Calibre manufacturing platform with NVIDIA’s cuLitho, a CUDA-X library that uses GPUs for computational lithography. The aim is to ease a runtime bottleneck as chipmakers adopt more complex patterning for leading-edge nodes, including High-NA EUV and curvilinear inverse lithography.
Reported gains start at 10x for key simulation workloads, with a headline result of 12x and verified data citing up to 17x for some tasks. Faster runs could let engineers explore more optimizations, find hotspots earlier, and improve process margins before tapeout.
Faster computational lithography may shorten development cycles for advanced AI chips, potentially affecting chip designers, foundries, mask makers, and eventually users of AI hardware. If it helps improve yields and process margins, more capable semiconductors could reach markets sooner, though benefits may concentrate among firms with access to leading-edge tools and GPU infrastructure. Broader effects on pricing, energy use, or access remain uncertain.