The GenAI era of Computational Lithography is coming

Computational lithography is used to optimize the quality of chips, and according to ASML's definition:

Computational lithography uses algorithmic models of the manufacturing process, calibrated with key data from our machines and from test wafers.

How important is computational lithography for the manufacturing of the most advanced computer chips? To find out, let's turn to public sources. Let's look at the recent public story from the US X-ray lithography startup Substrate. A large part of the article praises Google AlphaEvolve's AI agent, but there is a relevant passage on the subject:

We built a powerful computational lithography stack that includes processes like inverse lithography technology (ILT), which works backward from desired patterns to design optimal masks, and optical proximity correction (OPC), which adjusts mask designs to correct for distortions during printing.

It is interesting to note that one of the responsibilities of an AI Research Scientist at Substrate is to:

Implement surrogate models, physics-informed neural networks, or generative approaches for scientific problems

It appears that even in a field like photolithography, where precision and reproducibility matter a lot, the team at Substrate is venturing into physics-informed machine learning and generative AI to help solve inverse problems in a fast and efficient way, potentially achieved by using surrogate models in optimization loops.

Digging further into ASML's 2025 report, we read this:

In computational lithography, we are advancing optical proximity correction (OPC) technologies as part of our broader strategy for sustainability and operational excellence... 

We are increasingly using machine-learning techniques to further enhance the accuracy of models and reduce the computational time and cost... 

Machine learning and AI continue to enable these advanced techniques by delivering accuracy and speed.

If I were to predict the future of computational lithography, I would bet on the increased importance of this field in the coming years, as there seem to be low-hanging fruits that can be easily harvested with the use of AI in this space.