Synopsys has introduced the AgentEngineer suite, designed to automate chip verification, implementation, and manufacturing planning. The company currently has over 50 customer engagements and plans a general release by the end of the year.
Other companies are also integrating AI into hardware design. Cognichip offers an ACI platform for generating specifications and RTL, while Empyrean Technology reported that an AI agent reduced a circuit layout task from four weeks to one. OpenAI and Broadcom utilized AI models to develop the Jalapeño inference chip in nine months.
The use of AI in chip design raises concerns regarding intellectual property and patent infringement. Experts note that AI can inadvertently replicate existing designs by analyzing academic literature and patent databases, creating challenges in tracking the provenance of hardware components.
To mitigate risks, researchers suggest implementing shared repositories for encrypted design data to verify overlaps without exposing proprietary IP. Existing standards like IEEE 1735 may require updates to regulate how AI agents access and learn from design tools.
Despite the capabilities of AI, the semiconductor industry remains cautious due to the inability to patch fabricated silicon. Manufacturers currently limit AI involvement in architectural decisions, preferring to rely on licensed IP from established vendors that provide verification and standards compliance.
