AI's chipmaking frontier may face patent infringement hurdles as autonomous tools take over

AI is increasingly taking over tasks related to chipmaking, including design and verification. This shift raises concerns about potential patent infringement issues, as AI-generated designs may unintentionally infringe on existing patents. Companies such as Synopsys, CogniChip, and Empyrean Technology are already developing AI-driven solutions for chip design, with some already in use by clients. Even major AI labs are also involved, with OpenAI's models helping Broadcom develop their first custom inference chip.

The use of AI in chip design brings up questions about the role of human ingenuity in the industry. Intellectual property plays a crucial role in chip design, and it's unclear how AI-generated designs that infringe on existing patents will be addressed. While some existing tools can identify close copies of designs, AI-generated designs can be rewritten or run through synthesis tools, making it difficult to trace their origin.

Professor Domenec Forte from the University of Florida notes that AI could inadvertently copy other designs, as it works by ingesting large amounts of literature and constructing lookalike designs. This creates a "provenance problem" for chips, as it becomes challenging to document and explain the origin of AI-generated ideas and licensed IP.

International standards like IEEE 1735 could potentially help manage IP rights and manage the use of AI in chip design. However, Forte suggests that standards may need to be further defined to address the specific challenges posed by AI agents. Simon Moore, a professor at the University of Cambridge, emphasizes the importance of verification in chip design, as AI-generated test cases can help improve verification coverage.

However, allowing AI to make architectural decisions is still considered a no-brainer, with most manufacturers drawing the line at such decisions.