Chinese tech giant Baidu has unveiled two new AI chips, the Kunlun M100 and M300.

Launched at the Baidu World conference in Beijing this week, the company said the M100 supports large-scale inferencing workloads and will launch in early 2026, while the M300 has been developed for training and inferencing and is targeted for 2027.

Baidu also announced two ‘supernode’ solutions linking multiple chips together, dubbed the Tianchi 256 and Tianchi 512. Comprised of Baidu’s P800 chips, the 256 offering will be available in H1 2026, while the 512 version will be available from the second half of the year.

The Tianchi supernode is similar in concept to Nvidia’s GB200 NVL72 and Huawei’s CloudMatrix 384, or Atlas 950, the latter of which was announced in September of this year. Comprising 384 Huawei 910C chips, the Atlas 950 will be launched in the fourth quarter of 2026, supporting 8,192 Ascend chips per node.

The US government has continued to place restrictions on the export of advanced chips to China, and in recent months, the Chinese government has issued directives to limit the deployment of foreign chips by domestic companies.

In August, Chinese state media claimed that Nvidia’s H20 chips pose a national security concern to China, with reports also surfacing at the time that the Chinese government had previously expressed concerns over backdoor access being inserted into the semiconductors.

An article posted by a social media account associated with China’s state media said: “When a type of chip is neither environmentally friendly, nor advanced, nor safe, as consumers, we certainly have the option not to buy it.”

In March 2024, Robin Li, co-founder and CEO of Baidu, said that despite the restrictions placed on the purchase of cutting-edge AI chips by Washington, the user experience offered by Baidu “will not be compromised.”