Cognichip Secures $60 Million to Build AI That Designs the Next Generation of AI Chips

The most powerful AI models are built on advanced silicon. Now, a startup is betting that AI can be used to build that silicon itself. Cognichip is developing a deep learning system intended to work as a partner for human chip engineers, tackling a fundamental industry bottleneck: the sheer time and cost of designing a new processor.

Creating a state-of-the-art chip is a multi-year, billion-dollar endeavor. The design phase alone can stretch to two years before manufacturing even begins. Cognichip's founder and CEO, Faraj Aalaei, argues this timeline is unsustainable. "The market can shift completely, turning that massive investment into a loss," he said in an interview. His company aims to transplant the productivity gains seen in software development—where AI assists in writing code—into the semiconductor world.

The company announced a $60 million funding round led by Seligman Ventures. Notably, Intel's former CEO Lip-Bu Tan participated through Walden Catalyst Ventures and will join Cognichip's board. The round brings the startup's total funding to $93 million.

Cognichip's approach hinges on a model trained specifically on chip design data, not a general-purpose AI. Acquiring that data presented a significant hurdle, as chip blueprints are closely guarded intellectual property. The firm built proprietary datasets, used synthetic data, and established secure methods for partners to train models on their own designs. Where possible, it has turned to open-source architectures like RISC-V for development.

The startup faces established giants like Synopsys and Cadence, plus other funded entrants. While Cognichip claims its technology can slash development costs by over 75% and cut timelines in half, it has not yet publicly showcased a chip designed with its system or named its collaborating customers. For investors like Seligman's Umesh Padval, the current surge in AI infrastructure spending represents a historic opportunity. "If this is a super cycle for semiconductors," he said, "it's a super cycle for companies like this one."

Source: TechCrunch