One of the most complex tasks in modern engineering is chip design. However, a team in Intel’s massive campus in India is simplifying this task: “Gokul Subramaniam, Intel India president…says the semiconductor industry is embracing AI in ways that that few would have anticipated even a short time ago.”
Mr Subramaniam says that one can break down chip development into multiple stages – design, development, testing, validation and deployment. As one would expect, each stage has it workflow, numerous tools and specific people associated with it. AI agents can be used for each stage of this workflow: “We believe an agentic approach – thinking of them as sophisticated assistants – can help with tasks like generating code, running simulations, or assisting with placement (of transistors), all while engineers retain full oversight.”
So how will these AI agents be trained? Remember, unlike software development, where LLMs like OpenAI can be trained on billions of lines of publicly available code, the same can’t be done in the world of semiconductor engineering (where the code is closely guarded).
This is where Mr Subramaniam claims Intel has an advantage thanks to 50 years of its own historical data which can be used to train the AI agents. More specifically, this agent training is being done in Intel’s campus in India.
““We’re not building our own LLMs from the ground up,” Subramaniam says. “We’re using horizontal infrastructure – tools like LangChain or LangGraph – and layering our domain-specific knowledge on top.”…
In placement, for example, Subramaniam says an agent might suggest optimal layouts based on past designs, factoring in power, area, and timing constraints, while still leaving the final call to the human expert. In verification – a critical step before fabrication – agents simulate designs at scale, flagging potential errors that might take weeks to uncover.”
The TOI article extols other virtues of using AI agents for chip design has Mr Subramaniam claiming that this will lead to faster time-to-market, fewer errors and will “free up engineers to focus on the creative, not routine”. All of that being said this sounds similar to the tack CEOs in other industries are taking before scaling back their recruitment plans since the routine work will now be done by AI agents rather than freshly recruited graduates.
If you want to read our other published material, please visit https://marcellus.in/blog/
Note: The above material is neither investment research, nor financial advice. Marcellus does not seek payment for or business from this publication in any shape or form. The information provided is intended for educational purposes only. Marcellus Investment Managers is regulated by the Securities and Exchange Board of India (SEBI) and is also an FME (Non-Retail) with the International Financial Services Centres Authority (IFSCA) as a provider of Portfolio Management Services. Additionally, Marcellus is also registered with US Securities and Exchange Commission (“US SEC”) as an Investment Advisor.
