Cognichip wants AI to design the chips that power AI, and just raised $60M to try

Cognichip is taking a bold step forward in the world of semiconductor design by leveraging artificial intelligence to streamline the complex process of chip development. With a staggering $60 million raised in its latest funding round, the company aims to transform how engineers create the high-performance silicon chips critical for advancing AI technology. The challenge that Cognichip addresses is one that has plagued the chip industry for decades: the intricate and costly nature of chip design. Traditionally, the journey from concept to mass production can span three to five years, with the design phase alone taking up to two years. This lengthy process often results in wasted investments as market demands shift during development. Cognichip's CEO and founder, Faraj Aalaei, believes that AI can significantly reduce both the time and cost associated with chip design. According to Aalaei, the company’s innovative deep learning model can assist engineers by producing optimized design outputs based on specific guidance. "These systems have now become intelligent enough that by just guiding them and telling them what the result is that you want, it can actually produce beautiful code," he explained. Cognichip’s technology has the potential to cut chip development costs by over 75% and halve the typical timeline. Emerging from stealth mode last year, Cognichip has now accumulated a total of $93 million in funding since its inception in 2024. The latest investment round was led by Seligman Ventures, with significant contributions from Intel CEO Lip-Bu Tan, who will also join the company’s board alongside Umesh Padval, a managing partner at Seligman. Despite the promising outlook, Cognichip has yet to showcase a chip developed using its proprietary system and has kept details about its collaborators under wraps. The firm emphasizes its unique approach, utilizing a specialized model trained exclusively on chip design data, as opposed to general-purpose large language models (LLMs). This requires access to proprietary domain data, which is often tightly controlled in the industry. To overcome these challenges, Cognichip has built its own datasets, including synthetic data, and has secured licenses from partners. The company has also implemented protocols that allow chip manufacturers to train its models on their confidential data without the risk of exposure. In instances where proprietary data isn't available, Cognichip has relied on open-source alternatives. In an engaging demonstration, Cognichip invited electrical engineering students from San Jose State University to test its model during a hackathon, enabling them to design CPUs based on the open-source RISC-V architecture. Competing with established players like Synopsys and Cadence Design Systems, Cognichip is part of a growing wave of startups in the AI infrastructure market, with significant funding flowing into the sector. Padval noted that the current investment climate represents the largest influx of capital he has witnessed in four decades, indicating a strong future for companies like Cognichip in the semiconductor landscape.