Baidu Unveils Next-Generation AI Processors to Strengthen China’s Tech Independence

Baidu introduced two new artificial intelligence semiconductors on Thursday, describing them as cost-efficient and domestically controlled solutions designed to offer Chinese companies enhanced computing power. The launch comes amid intensifying U.S.–China tensions, which have resulted in restrictions on advanced American AI chip exports to China, driving local firms to accelerate the development of homegrown processors.

At its annual Baidu World technology conference, the company revealed that the M100 chip, designed primarily for AI inference, will be released in early 2026, while the M300—capable of handling both training and inference tasks—is expected to debut in early 2027. In AI development, training involves enabling models to learn from vast datasets, whereas inference applies those trained models to generate predictions and respond to user inputs.

Having produced proprietary chips since 2011, Baidu also introduced two “supernode” systems, which integrate multiple chips through advanced networking technology to overcome the performance constraints of individual processors. Huawei has launched a comparable system, the CloudMatrix 384, consisting of 384 Ascend 910C chips, which analysts view as surpassing Nvidia’s GB200 NVL72 in power. Huawei further disclosed plans in September to roll out even more advanced supernode systems in the future.

Baidu’s upcoming Tianchi 256 system, composed of 256 P800 chips, is scheduled for release in the first half of next year, followed by a larger version with 512 chips in the latter half. The company also introduced an upgraded edition of its Ernie large language model, highlighting its enhanced capabilities in text understanding as well as image and video processing.