Cutting-edge silicon chips are accelerating the development of artificial intelligence. So, can AI return the favor?
CogniChip builds deep learning models to help engineers design new computer chips. The problem the company is solving is one the industry has had for decades. Chip design is extremely complex, prohibitively expensive, and time-consuming. Advanced chips take three to five years to go from concept to mass production. The design phase alone can take as long as two years before physical layout begins. Consider that the latest Nvidia GPU, Blackwell, contains 104 billion transistors. This is quite a large amount when put side by side.
During the time it takes to develop a new chip, the market can change and all that investment could be wasted, said Faraj Aaraei, CEO and founder of CogniChip. Aalaei’s goal is to bring the kinds of AI tools that software engineers have been using to speed up their work to the semiconductor design space.
“These systems are now intelligent enough that you can actually generate beautiful code by simply guiding the system and telling it the outcome you want,” Aalaei told TechCrunch.
He said his company’s technology can reduce chip development costs by more than 75% and cut timelines by more than half.
The company came out of stealth last year and on Wednesday announced it had raised $60 million in new funding led by Seligman Ventures. The funding included notable participation from Intel CEO Lip Vu Tan, who invested through his venture firm Walden Catalyst Ventures and will join CogniChip’s board of directors. Umesh Padval, managing partner of Seligman, will also join the board of directors. Since its founding in 2024, Cognichip has now raised a total of $93 million.
Still, Cognichip has not yet been able to point to any new chips designed with its system, nor has it identified any of the customers it says it has been working with since September.
tech crunch event
San Francisco, California
|
October 13-15, 2026
The company says the advantage is that it uses proprietary models trained on chip design data, rather than starting with a generic LLM. This required access to domain-specific training data, which was no small task. Unlike software developers, who openly share vast amounts of code, chip designers are fiercely protective of their intellectual property and have little access to the open source treasure troves that typically train their AI coding assistants.
Cognichip had to develop its own datasets, including synthetic data, and license data from partners. The company has also developed a procedure that allows chip makers to safely train CogniChip models on that data without having to make their proprietary data public.
When proprietary data is not available, CogniChip has relied on open source alternative data. In a demonstration last year, CogniChip invited San Jose State University electrical engineering students to try out the model at a hackathon. Using this model, the team was able to design a CPU based on the RISC-V open-source chip architecture, a freely available design that anyone can build.
Cognichip competes with established players such as Synopsys and Cadence Design Systems, as well as well-funded startups such as ChipAgentsAI, which closed a $74 million expansion Series A in February, and Ricursive, which raised a $300 million Series A round in January.
Padval said the current influx of capital into AI infrastructure is the largest investment in 40 years.
“If this is a supercycle for semiconductors and hardware, it’s a supercycle for companies like (CogniChip),” he said.
