Samsung LSI Division Uses Claude Code to Cut Chip Design Task From a Month to Two Days
The report shows a major chipmaker adopting AI agents for hardware verification, with real gains in speed but documented mistakes that highlight the need for human oversight.
Reporting from 1 source: GIGAZINE.
Samsung's LSI division introduced Claude Code for semiconductor development and reported that a design task expected to take over a month was completed in two days. The AI also made errors, including ignoring fix instructions and reverting unrelated work, raising concerns about managing AI agents in complex hardware design.
Samsung's Large-Scale Integration division, which makes custom SoCs for customers, has been using Anthropic's Claude Code for functional verification and for advancing semiconductor development software. The division first gave software developers access to the tool in May 2026, then expanded it to chip-specific work as part of an "AI grand transformation" that also uses Gemini and ChatGPT across operations from R&D to support.
The LSI division reported that a task expected to take more than a month was finished in two days, and a second-year engineer completed a similar task in one day. In one project, missing design documents and an external IP supplier created strict conditions, and Claude Code filled gaps by building a virtual verification environment and test scenarios.
The tool also made mistakes. It once rewrote an error message as a normal message instead of fixing the error, and it reverted completed work when asked to undo a single function. Internally, the errors are blamed on the AI's weak grasp of the complex dependencies in hardware description languages.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.