Cognichip raises $60M for a chip-design AI model it says cuts costs by over 75%

Startup building a deep learning model to work alongside engineers designing new computer chips

A deep-learning model trained on chip design data, not a general LLM, can work beside engineers to cut chip development cost and time

$60M new round led by Seligman Ventures; $93M raised in total since founding in 2024

CEO and founder Faraj Aalaei started Cognichip in 2024 to bring AI tools like those used by software engineers into semiconductor design. The company came out of stealth in 2025.

Chip designers guard their IP closely, so unlike software there is no open trove of training data. Cognichip built its own datasets, including synthetic data, licensed data from partners and developed ways for chipmakers to train on their own data securely. In a 2025 demo, San Jose State electrical engineering students used the model in a hackathon to design RISC-V CPUs.

Chip design is enormously complex, expensive and slow, taking up to two years before physical layout

The market can change during the three to five years a chip takes to reach production

A domain-specific model avoids relying on a general LLM that lacks chip-design data

An investor called the flow of capital into AI infrastructure the largest in 40 years

Cognichip's value rests on claims (over 75% lower cost, more than half the time) that it has not yet proven on a finished chip.

As of 2026-04-01 Cognichip had $93M in total funding and an Intel CEO on its board, but no publicly named customers or chip designed with its system. Its cost and time claims are the CEO's own.