News

About TrendForce News

TrendForce News operates independently from our research team, curating key semiconductor and tech updates to support timely, informed decisions.

[News] China Reportedly Unveils World’s First Fully Automated AI Chip Design System amid U.S. Curbs


2025-06-13 Semiconductors editor

While the U.S. aims to curb China’s chip industry with EDA export restrictions, the Chinese Academy of Sciences recently launched “QiMeng,” the world’s first fully AI-powered system for automated processor chip design, according to EE Times China.

The report suggests that in terms of CPU design, the system has produced the world’s first fully automated CPU chip, “QiMeng No.1,” which is capable of completing the entire front-end design of a 32-bit RISC-V CPU within just five hours. Notably, with tape-out finished, the chip boasts Intel 486-level performance and packs over four million logic gates, according to EE Times China.

“QiMeng No. 2,” its upgraded version, is reportedly the world’s first fully automated superscalar processor core, delivering performance on par with the ARM Cortex-A53, EE Times China suggests. With a scale of 17 million logic gates, it has reportedly designed to meet the demands of high-performance edge computing.

Notably, the research team plans to push QiMeng into advanced nodes like 3nm and deepen collaboration with foundries to speed up its rollout, according to EE Times China.

According to the paper summary on arXiv, QiMeng is structured in three layers. At its foundation lies the Large Processor Chip Model (LPCM), a specialized chip model designed to overcome hurdles like scarce data, complex knowledge, and accuracy assurance. Building on this, the middle layer features intelligent agents that automatically design both the hardware and software of processor chips. Early results show that parts of QiMeng are already proving effective in real-world applications.

Challenges ahead

While QiMeng hints at reduced reliance on EDA giants like Synopsys and Cadence amid rising geopolitical tensions, EE Times China notes that commercialization won’t be easy. The system still needs to prove its stability at scale and build a robust support ecosystem before it can go mainstream, as the report points out.

Therefore, Tech Wire Asia notes that QiMeng’s early applications are likely to focus on domain-specific processors—tailored chips optimized for targeted tasks, where AI-driven design could make the biggest impact.

As highlighted by Tech Wire Asia, designing chips is not just about making them work—it also requires fine-tuning for manufacturing yield, heat control, electromagnetic safety, and long-term reliability, often across decades. Current AI systems still struggle to juggle all these demands at once, the report adds.

Read more

(Photo credit: AMEC)

Please note that this article cites information from EE Times China, arXiv and Tech Wire Asia.


Get in touch with us