Cognichip raises $60M to teach AI to design chips for AI

Startup Cognichip has raised $60 million to develop technology that uses artificial intelligence to design the chips themselves, the very chips modern AI cannot exist without. The company says its approach will cut semiconductor development costs by more than 75% and timelines by more than half.

Those are radical numbers for an industry where a new chip traditionally takes years and costs billions of dollars. Nvidia, Apple, Google and other players pour enormous resources into designing, verifying and testing each new generation of processors. Cognichip wants to automate a significant part of that process with AI agents that can make architectural decisions, optimize circuits and reduce the number of iterations between engineering teams.

In essence, the idea is an AI copilot for chip designers or, in a more ambitious scenario, full automation of certain stages of development.

The startup is part of the fast-growing EDA (Electronic Design Automation) segment, tools that automate electronics design. Established players such as Synopsys and Cadence are already adding machine learning to their platforms, but Cognichip is betting on an AI-native approach built from scratch rather than a layer on top of legacy systems.

The loop is notable: the shortage of compute for training AI models is directly tied to how quickly the industry can design new chips. If Cognichip delivers on its promises, it will speed up not only hardware production but also the development of the AI that consumes that hardware.

Claims of radical cost cuts in the semiconductor industry surface regularly, and the startup is still a long way from commercial results. Even so, the $60 million round signals serious investor interest in closing the loop: using AI to build the infrastructure AI runs on.