Cognichip just raised $60 million to build deep learning models that help engineers design computer chips faster and cheaper. The funding round was led by Seligman Ventures, with Intel CEO Lip-Bu Tan personally participating and joining the company’s board, TechCrunch AI reports.
This matters because chip design is one of the last major engineering disciplines that AI hasn’t meaningfully transformed yet.
The problem Cognichip is attacking
Advanced chips take three to five years from conception to mass production. The design phase alone can eat up two years before physical layout even starts. Nvidia’s latest Blackwell GPUs pack 104 billion transistors. That’s an insane amount of complexity to manage by hand.
Cognichip CEO Faraj Aalaei says by the time a new chip reaches production, the market may have already moved on, making the entire investment worthless.
“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,” Aalaei told TechCrunch.
The company claims its technology can:
- Cut chip development costs by more than 75%
- Reduce timelines by more than half
What makes this approach different
Cognichip built its own model trained specifically on chip design data, rather than fine-tuning a general-purpose LLM. That’s a harder path, but potentially a more effective one.
The challenge? Getting training data. Unlike software, where open source code is everywhere, chip designers guard their IP closely. Cognichip had to develop its own datasets, including synthetic data, and license data from partners. They’ve also built secure procedures so chipmakers can train models on their proprietary designs without exposing them.
In a demo last year, electrical engineering students at San Jose State University used the model in a hackathon to design CPUs based on the RISC-V open source architecture. Promising, but still early.
The competitive landscape is heating up
Cognichip isn’t alone in this space. According to TechCrunch AI, it’s going up against:
- Incumbents: Synopsys and Cadence Design Systems, who dominate chip design tools today
- ChipAgents: closed a $74 million extended Series A in February
- Ricursive: raised a massive $300 million Series A in January
With $93 million raised total since its 2024 founding, Cognichip is well-funded but not the biggest spender in the room.
What’s still missing
Cognichip can’t yet point to a finished chip designed with its system. It also didn’t disclose any of the customers it says it’s been collaborating with since September. That’s a notable gap. Cutting design time in half is a bold claim that needs real-world proof.
Why this trend matters
Seligman managing partner Umesh Padval, who’s also joining Cognichip’s board, called the current flood of capital into AI infrastructure the largest he’s seen in 40 years of investing. “If it’s a super cycle for semiconductors and hardware, it’s a super cycle for companies like Cognichip,” he said.
What stands out here is the recursive logic: AI designing the chips that run AI. If tools like Cognichip’s deliver on their promises, they could dramatically accelerate the hardware innovation cycle. Faster chip design means faster chips, which means better AI, which means even better chip design tools. That feedback loop is exactly why investors are pouring hundreds of millions into this space.
For now, the proof will come when the first Cognichip-designed silicon actually ships. More details are available through TechCrunch AI’s original coverage.