Cognichip Revolutionizes Chip Design with $60M and AI

Cognichip and Its Ambitious Fundraising

Cognichip, an innovative startup in the semiconductor field, has recently raised $60 million to enable artificial intelligence to take charge of electronic chip design. This fundraising round was led by Seligman Ventures, with notable participation from Lip-Bu Tan, CEO of Intel, who is joining Cognichip's board of directors. Umesh Padval, managing partner at Seligman, has also been appointed to the board. Since its founding in 2024, the startup has accumulated a total of $93 million in funding.

Cognichip's approach is to let AI give back to the ultra-high-performance silicon chips that have enabled its rise. Indeed, these chips have been essential for the development of AI, and it is now time for AI to contribute to their design.

Cognichip's Ambitious Project

Cognichip's mission is to develop a deep learning model aimed at assisting engineers in chip design. This process is traditionally long and complex, requiring between three to five years from the initial sketches to mass production. The design phase alone can stretch over two years before manufacturing even begins.

With next-generation chip models, such as Nvidia's Blackwell GPU, which boasts 104 billion transistors, the complexity is only increasing. Faraj Aalaei, CEO and founder of Cognichip, emphasizes that this timeline is problematic because the market often evolves faster than the chips themselves, rendering some products obsolete even before their release.

Cognichip's idea is to import into the semiconductor sector the methods already employed by software developers, using AI to accelerate the process. According to Aalaei, current systems are advanced enough to produce high-quality code, provided they are given clear instructions. Applied to chip design, this approach could transform the industry.

Cognichip claims it can reduce development costs by over 75% and halve timelines thanks to its model.

Cognichip's Concrete Advances

For now, Cognichip remains discreet about its concrete achievements. No chip designed with its system has yet been presented, and the startup has not revealed the identities of its clients, although collaborations have been mentioned since September. However, Cognichip has a major asset: a proprietary model specifically trained on chip design data, which sets it apart from generalist models.

Access to this data is particularly challenging. Unlike software developers who widely share their code, semiconductor designers fiercely protect their work. To address this lack of accessible data, Cognichip has created its own datasets, combining synthetic data and licensed content. It has also implemented mechanisms that allow manufacturers to train models on their internal data without exposing it.

In parallel, Cognichip is leveraging open-source alternatives. During a hackathon organized last year, electrical engineering students from San José State University were able to test its model, designing processors based on the RISC-V architecture, an open standard that can be adapted by anyone.

Cognichip's Competition

Cognichip is not alone in this market. It faces established giants like Synopsys and Cadence Design Systems. Synopsys offers comprehensive solutions covering the entire lifecycle of a component, from logical design to simulation and verification, allowing for testing a chip's behavior in various scenarios before manufacturing.

Cadence Design Systems takes a slightly different approach, also relying on three pillars: design, simulation, and verification. Its tools use machine learning to automatically optimize designs, reduce errors, and accelerate timelines.

Emerging Rivals

A new generation of ambitious startups is also beginning to emerge. ChipAgents is developing an agent-based AI platform to accelerate the chip design and verification process. In February, it closed a $74 million funding round, led by Matter Venture Partners.

Ricursive Intelligence is another ambitious challenger, founded by researchers who worked on projects like AlphaChip at Google. It aims for fully autonomous design guided by reinforcement learning techniques and recursive AI. In January, the startup raised $300 million, valuing it at around $4 billion just two months after its launch.

Cognichip must therefore be wary of this growing competition.