Chinese Baidu Unveils Two New AI Chips
The company introduced the M100, engineered to optimize inference throughput in mixture-of-experts architectures, alongside the M300, calibrated for training colossal multimodal systems spanning trillions of parameters, according to media reporting.
Shen Dou, Baidu's executive vice-president and cloud division chief, positioned both chips as instrumental to Beijing's self-reliance agenda. The M100, arriving early 2026, and the M300, targeted for 2027, will deliver "powerful, low-cost, and controllable AI computing power" to support China's drive for AI self-sufficiency," he stated.
Baidu projects exceeding 50% efficiency gains across AI infrastructure deployments, with the Tianchi256 stack slated for first-half 2026 release. An expanded iteration, Tianchi512—integrating 512 processors—will follow in the second half of 2026.
By 2030, the company envisions constructing a supernode architecture capable of supporting "millions" of chips.
Baidu founder and CEO Robin Li Yanhong assailed the AI sector's structural imbalances, characterizing the industry as "very unhealthy" due to a system that disproportionately rewards developers of chips and foundational AI models."
The competitive landscape intensified following recent White House action. Spokesperson Karoline Leavitt announced the U.S. will not allow Nvidia to sell its most advanced AI chip to China for now." President Trump separately declared that cutting-edge semiconductors from America's most valuable corporation would be restricted to domestic enterprises and prohibited from sale to Beijing and allied nations.
Li disclosed that approximately half of Baidu's newly generated software code now originates from its proprietary AI coding platform, Baidu Comate. "But 50% is not enough; I hope that in the future we can reach 80 or 90%," he said."
Separately Thursday, Baidu premiered its latest natively omni-modal foundational model, ERNIE 5.0, via X, the social media platform. The architecture enables omni-modal comprehension, creative composition, instruction compliance, and ancillary advanced functions.
Legal Disclaimer:
EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
