Samsung reports that it has introduced Claude Code and 'reduced a design process that used to take a month to two days.'
It has been revealed that Samsung has adopted Anthropic's AI coding tool 'Claude Code,' significantly shortening the time required for some semiconductor design and verification processes.
[단독] 성과 - 조선비즈
Samsung says Claude Code can cut chip design work from weeks to days, but it still makes serious mistakes | TechSpot
According to an interview with Chosun Biz, Samsung's Large-Scale Integrated Circuit (LSI) division, which has implemented Claude Code, is using it for functional verification of custom SoCs for customers and for advanced development of semiconductor development software. Samsung has officially announced an 'AI transformation' in May 2026, which involves prioritizing the release of Claude Code to software development personnel, then expanding its use to specialized semiconductor development tasks, and utilizing other generative AIs such as Gemini and ChatGPT to integrate AI into all operations from research and development to production, marketing, and support.
The LSI division has reported that a task that was expected to take more than a month was completed in just two days. There have also been cases where engineers in their second year at the company completed development work that could have taken more than a month in just one day.
In one project, the client company requested a new semiconductor structure, and the use of design assets (IP) from an external company made the development conditions very demanding. Some design documents were missing, and the necessary verification documents were not completed by the scheduled time, so Samsung used Claude Code to fill in the gaps. By inputting the information they obtained into Claude Code, they were able to find the necessary information and create a virtual verification environment and test scenarios.
However, Claude Code also made mistakes. In one case, it ignored instructions to correct an error and instead rewrote the error message into a normal message. In another instance, when asked to revert only a specific function, it also reverted other tasks that had been completed earlier. The company has determined that this was a problem caused by the AI's inability to fully grasp the complex dependencies of the hardware design language.
A semiconductor industry insider interviewed by Chosun Biz said, 'AI agents are fast, but if not properly managed, they could lead to major accidents. Ultimately, development will progress in a direction where the number of manual steps previously performed by humans is reduced to shorten the overall development time, allowing engineers to focus on goal setting and final verification.'
Related Posts:
in AI, Posted by log1p_kr