China’s Progress in 5nm AI Chips: Domestic Companies Announce Two New Solutions for Model Training and AI PC Applications

China is making significant strides toward the development of advanced 5nm AI GPUs. According to recent reports, local companies are preparing to launch new chips that promise to enhance performance levels in the field of artificial intelligence.

Collaboration Between Anfu Technology and Xiangdi: The Introduction of New 5nm AI Chips

Beijing is determined to produce homegrown solutions that minimize dependence on Western technologies while also fostering a robust internal tech ecosystem. A report from MyDrivers highlights that China’s Anfu Technology is teaming up with Xiangdi to create a new class of ‘Fuxi’ GPUs. These chips are touted to deliver exceptional performance, particularly in handling parallel computing tasks. Currently, two models are under development to serve both the AI training and personal computer markets, featuring cutting-edge 5nm technology.

The upcoming Fuxi GPUs are set to provide impressive capabilities, offering up to 160 TFLOPS of FP32 computational power sourced from a 5nm architecture. This development is particularly noteworthy since local semiconductor manufacturers have struggled to progress beyond the 7nm threshold. While there have been rumors regarding the presence of TSMC’s 5nm technology in the country, it is not confirmed that Anfu Technology is utilizing these nodes. Nevertheless, these advancements signify a pivotal moment for the Chinese AI sector, reflecting its commitment to achieving extraordinary computational enhancements.

Rapidus semiconductor facility in Japan aiming for 2nm chip mass production by 2027 amid competition from TSMC and Samsung

Among the new developments, the Fuxi A0 chip is designed specifically for rendering tasks, while the Fuxi B0 chip targets AI applications and features an integrated Neural Processing Unit (NPU).The B0 model is notably compatible with mainstream AI frameworks, including DeepSeek R1, making it a promising candidate for practical implementations. Meanwhile, the A0 chip is tailored for tasks within the emerging ‘AI PC’ segment, focusing on optimized deployment of AI models at the end-user level.

While comprehensive details about the new AI solutions from Anfu Technology and Xiangdi remain limited, the rapid evolution of China’s computational capabilities is striking. This progress is largely fueled by the government’s strategy to decrease reliance on NVIDIA’s AI products, positioning domestic giants like Huawei and Cambricon as viable alternatives in the semiconductor sphere.

Source&Images