Attempts to use artificial intelligence to reshape semiconductor design are gathering pace. The move aims to shorten design timelines and cut costs that under existing structures can take years.

IT outlet TechCrunch reported on April 1 that Cognichip, which develops AI models for semiconductor design, recently raised $60 million in new funding. The round was led by Seligman Ventures. Intel CEO Lip-Bu Tan (립부 탄) participated through his venture fund and also joined the board. Cognichip's total funding rose to $93 million.

The core issue Cognichip is targeting is the time and cost of semiconductor design. Advanced chips typically take 3 to 5 years from concept design to mass production, and the pre-physical layout design phase alone can take up to 2 years. Design complexity is also rising sharply. For example, Nvidia's latest GPU architecture Blackwell includes about 104 billion transistors, greatly increasing the difficulty of design and optimisation.

Cognichip's strategy is to address this bottleneck with AI. Paraj Alaei (파라즈 알라에이), the company's co-founder and CEO, said, "AI systems have advanced to the point where they can generate high-quality code if you present the desired outcome," adding, "It can dramatically boost productivity in semiconductor design as well." The company claims its technology can cut chip development costs by more than 75 percent and shorten development time by more than half.

The technical differentiator lies in a dedicated design model rather than a general-purpose large language model. Because semiconductor design data is tightly restricted from external disclosure due to intellectual property protection, Cognichip built its model by combining synthetic data, licensed data and its own secure training protocol. It said this enables training without exposing customers' sensitive data externally.

In areas where data is hard to secure, it also used the open-source ecosystem. The company demonstrated a CPU design model based on the open-source chip architecture RISC-V at a hackathon for San Jose State University students, showing the feasibility of the technology.

It has not yet fully reached the commercialisation stage. Cognichip has not disclosed any new chips designed with its system, and it has not named any customers it is working with. Some assessments also say results will need to be validated because the investment was made during development rather than after technical verification.

The competitive landscape is also tough. Cognichip is competing not only with Synopsys and Cadence Design Systems, established leaders in the electronic design automation market, but also with AI-based startups. Alpha Design AI and ChipAgentsAI have also raised large investments recently and entered the market.

Investors see this trend as part of an AI infrastructure supercycle. Umesh Padval (우메시 파드발), managing partner at Seligman Ventures, said, "The capital flowing into AI infrastructure is the largest I have seen in my 40-year investing career," adding, "The supercycle in semiconductors and hardware will also be an opportunity for companies like Cognichip."

As AI expands from software development into semiconductor design, attention is on whether Cognichip's efforts can lead to structural change in the industry. Demonstration of the technology, commercialisation and performance against established players remain tasks ahead.