The United States has once again tightened its grip on technology exports to China by restricting the sale of EDA tools, which are crucial for semiconductor design. This move significantly impacts companies like Xiaomi, limiting them to using TSMC’s 3nm process and obstructing their efforts to develop in-house 2nm SoCs. In anticipation of future restrictions, the Chinese Academy of Sciences (CAS) has launched an innovative chip design system called QiMeng, harnessing artificial intelligence to streamline semiconductor development and reduce the need for human intervention.
QiMeng, meaning ‘enlightenment,’ has already demonstrated its potential by designing two processors. However, these processors mirror technology from generations past. The newly introduced autonomous system is capable of achieving results comparable to those created by human experts, significantly speeding up processes like designing chips for self-driving cars. While this task might take weeks for a team of experts, QiMeng can accomplish it in just days.
The system is built on three interconnected layers: the foundation consists of a domain-specific large processor chip model, the middle houses a hardware and software design agent, and the top features various processor chip design applications. With these components, Chinese researchers have successfully created two processors: the QiMeng-CPU-v1, similar to Intel’s 486 chip from over 36 years ago, and the QiMeng-CPU-v2, akin to ARM’s Cortex-A53 design.
A recent research paper highlighted on GitHub discusses the challenges inherent in semiconductor design, such as the absence of advanced fabrication technology and resources, which hinder QiMeng’s full potential. Despite these hurdles, the system aims to improve efficiency, cut costs, and shorten development cycles. However, without cutting-edge machinery, China remains reliant on older lithography techniques.
In related efforts, Huawei’s partner SiCarrier is developing machinery that may rival Dutch giant ASML. Earlier, SiCarrier sought a $2.8 billion funding round to bridge the technological gap. Additionally, Huawei has successfully developed 14nm EDA tools to mass-produce the 7nm Kirin 9020. To remain competitive with the U.S., significant investment in research will be vital to leveling the technological playing field.
