Chips are an important component of the AI industry. However, new chips do not hit the market at the same speed as new AI models and products.
Cognichip has the noble goal of creating basic AI models that will help bring new chips to the market faster.
San Francisco-based Cognichip is working to build a physics-based basic AI model that semiconductor companies can use to speed up the development process for new chips. The company calls this approach “artificial chip intelligence” and hopes to speed up chip production times by 50% and reduce the associated costs.
This ambitious idea comes from semiconductor industry veteran Faraj Aalaei, and worked for a variety of companies, including Fujitsu Network Communications and Centrium Communications.
Aalaei told TechCrunch that the story of his company’s origins begins in 2015. At the time, Aalaei was a member of the Silicon Valley Leadership Group.
He was worried about what was going on in the semiconductor industry. He presented to the group about venture capital companies investing in semiconductor companies, which peaked in 200 deals a year in 2000 and fell to one or two per year by 2015.
“I essentially warned other CEOs that this is not good for us,” Araei said. “It’s not good for the American semiconductor industry. All we have to do is fundamentally change it. If that trend continues, we lose competitiveness. We lose energy to bring new ideas to the table.”
Given how long it will take for new chips to hit the market, it’s not surprising that these companies aren’t attracting venture investors, he said. He then sat on the idea for almost ten years.
He found Candou Ventures in 2016 and through its funds he was able to see the rise of AI startups. I decided to launch Cognichip in 2024 when I realized that advances in generative AI could now be used to solve some of the existing challenges of the semiconductor industry.
Cognichip has since opened in stealth and has started building with a team of AI experts from locations such as Stanford, Google, and MIT. Aalaei said it would take at least several years to build the model into “ultimate performance,” but said it should be able to help businesses before they reach that goal.
“When you get to that point, you’re building a system that actually works like an expert engineer,” Aalaei says. “Once you achieve that vision, you can actually get the same task, with just a fraction of people, in a much shorter time.”
Cognichip is now emerging from stealth with $33 million seed funds in the round co-led by Lux Capital and Mayfield, responsible for participation from FPV and Candou Ventures.
Mayfield’s managing partner Navin Chaddha told TechCrunch that when he was introduced to Aalaei, he felt they were “cut from the same fabric.” Most of the work in the semiconductor industry is still done by humans. He said that Timing thinks it’s right to incorporate AI into the mix.
“This is a big problem, and the solution this company offers is not vitamins in the semiconductor industry, not painkillers,” says Chaddha. “If humans aren’t doing their jobs, can AI do that in places where there is a lack of talent?
Aalaei said Cognichip hopes to help Cognichip democratize access to its buildings so that more semiconductor companies can start and invest in land. Also, making access easier means smaller companies can build more concrete chips for professional and small models, he said.
All of this depends on when a company can reach artificial chip intelligence or when.
“What we’re doing isn’t progressive change,” Araei said. “We’re not building any (automating electronic design) tools. We’re not trying to tweak the process a bit. We’re trying to set new goals for the industry and bring about major changes.”