Baidu Makes a Move with New Artificial Intelligence Chips

Baidu Introduces Two New Chips for Artificial Intelligence

Baidu introduced two new semiconductor products for artificial intelligence on Thursday, stating that these products could provide Chinese companies with powerful, low-cost, and locally controlled computing power.

The increasing tensions between the United States and China have led to restrictions on the export of advanced artificial intelligence chips from U.S. firms to Chinese companies. This situation has prompted many firms to develop their own processors or look for local alternatives.

Baidu unveiled a decision-based chip named M100, which is planned to be released in early 2026, at the annual Baidu World technology conference. Additionally, it was announced that the M300 chip, capable of training and inference, will be introduced in early 2027.

Training involves building artificial intelligence models by learning patterns from large datasets, while inference uses these models to make predictions and process user requests.

Baidu has been developing its own chips since 2011. The company also announced new supernode products aimed at compensating for limitations in individual chip performance by combining several chips.

Huawei developed a product named CloudMatrix 384, which industry observers consider more powerful than Nvidia's advanced system-level product GB200 NVL72. Additionally, it announced plans to release more powerful supernode products in September.

Baidu's product named Tianchi 256 will consist of 256 P800 chips and is set to be launched in the first half of next year. A more powerful version with 512 chips is planned to be introduced in the second half.

Furthermore, Baidu announced a new version of its Ernie large language model, which excels not only in processing text but also in image and video analysis.