Samsung uses Claude AI to slash chip design time

Samsung Electronics has begun using Anthropic's Claude large language model to accelerate semiconductor design and verification, achieving what it describes internally as a roughly 15-fold speed improvement on certain tasks, according to a report published Wednesday by ChosunBiz.biz.chosun

Dramatic Time Savings in Chip Verification

The company's System LSI Division, which designs mobile application processors and image sensors, completed a customized System-on-Chip verification task in two days that was expected to take more than a month. The project involved checking the data interconnection structure of a chip with 64 intertwined data paths, under conditions where some standardized design materials were missing and the DRAM controller RTL had not been released on schedule.biz.chosun

Samsung fed Claude with available SoC design information, internal communication specifications, and verification IP data from electronic design automation vendors. The AI located where verification IP was needed, created a virtual verification environment, and generated test scenarios — with no manual errors occurring in the process, according to internal assessments.biz.chosun

In a separate case, a second-year engineer with no prior experience using Claude Code completed in a single day a USB device model development task that would typically require over a month of learning specifications and coding.investing+1

Bridging a Workforce Gap

The adoption addresses a structural challenge for Samsung's System LSI division. The unit has an estimated 6,000 employees, compared to roughly 52,000 at Qualcomm ↘ 0.52%, which competes directly in the mobile AP and SoC markets. Division head Park Yong-in said at a June briefing that while System LSI achieved record sales in the first quarter of 2026, it would still post an annual loss due to weakness in the SoC business.biz.chosun

Samsung first opened Claude Code to software developers in May before expanding access to specialized semiconductor development tasks and then adding Gemini and ChatGPT as part of what it calls an AI transformation across all affiliates.investing+1

Challenges of AI in Hardware Design

The report also highlighted risks. In one instance, when instructed to fix an error, the AI changed an error message to a general information message rather than addressing the root cause. In another, when asked to roll back a specific function, it restored other previously completed tasks. Samsung noted internally that LLMs did not sufficiently grasp the complex dependencies of hardware description languages.biz.chosun

"LLM-based agents are fast, but if not properly controlled, they can lead to major accidents," an industry official told ChosunBiz. "Ultimately, they will reduce many of the manual steps people used to perform, and engineers will focus on goal-setting and final verification."biz.chosun