The development of artificial intelligence has been dramatically accelerated by the most advanced silicon chips. Now, can AI aid in the creation of new chips? Cognichip is poised to do just that by building a deep learning model to assist engineers in designing new computer chips. The company is tackling the longstanding challenges in the industry: complexity, cost, and lengthy timelines in chip design. Designing advanced chips can take three to five years, with the design phase alone extending up to two years before physical layout begins. For example, Nvidia’s latest GPUs, Blackwell, comprise 104 billion transistors.
Faraj Aalaei, CEO and founder of Cognichip, emphasizes the urgency of this innovation, explaining how the market dynamics can render investments in lengthy chip development obsolete. Aalaei aims to introduce AI tools similar to those used by software engineers into the semiconductor design process, accelerating it significantly. According to Aalaei, the company’s technology can reduce development costs by over 75% and cut the timeline by more than half.
Having emerged from stealth last year, Cognichip recently raised $60 million in new funding led by Seligman Ventures, with notable contributions from Intel CEO Lip-Bu Tan via Walden Catalyst Ventures. This brings Cognichip’s total funding to $93 million since its 2024 inception. Yet, despite its cutting-edge technology, Cognichip hasn’t yet disclosed a new chip developed with its system nor identified any collaborating customers.
Cognichip claims its edge lies in its proprietary model trained on chip design data, rather than using a general-purpose language model. Acquiring domain-specific training data has been challenging due to chip designers’ reluctance to share their closely-guarded intellectual property. Consequently, Cognichip has created its own datasets and secured data licenses from partners. The company has also developed secure methods for chipmakers to train Cognichip’s models on their proprietary data.
Where proprietary data isn’t accessible, Cognichip has utilized open-source alternatives. In a demonstration last year, they engaged electrical engineering students at San Jose State University in a hackathon, using the model to design CPUs with the open-source RISC-V architecture.
Facing competition from established companies like Synopsys and Cadence Design Systems, as well as startups like Alpha Design AI and ChipAgentsAI, Cognichip is navigating a landscape flooded with investment in AI infrastructure. Umesh Padval from Seligman highlights this influx of capital as unprecedented in his 40 years of investment experience, predicting it marks a super cycle for semiconductors and companies like Cognichip.