Baidu is expanding its artificial intelligence chip strategy as China accelerates efforts to reduce its reliance on Nvidia technology.
The country’s tech giants continue to report strong demand for AI computing, but access to advanced chips has tightened due to export restrictions and global shortages.
This environment has pushed Chinese platforms to build domestic hardware pipelines that can support training and deployment of large AI models at scale.
Although Baidu is widely recognised for its search engine, the company has steadily repositioned its business around autonomous driving, cloud computing and AI systems.
Its semiconductor arm Kunlunxin is now one of the key pieces in China’s attempt to build more control over its AI infrastructure.
The company is increasing investment, expanding product plans and securing more orders as AI hardware becomes one of China’s most competitive technology frontiers.
Kunlun roadmap sets up long-term strategy
Baidu has introduced a detailed five-year roadmap for Kunlunxin, beginning with the M100 chip expected in 2026 and the M300 due in 2027.
These processors are intended to support training and inference tasks for Baidu’s ERNIE models, which currently run on a mix of self-developed chips and Nvidia hardware inside the company’s data centres.
The semiconductor unit is gaining more commercial traction. Kunlunxin has received orders from suppliers serving China Mobile, one of the country’s largest mobile operators.
Baidu also sells its chips to companies building data centres and offers computing capacity through its cloud services.
Experts have noted that Baidu is increasingly positioning itself as a full stack AI technology company, with infrastructure that combines chips, servers, data centres, models and applications.
Forecasts suggest that demand for domestic hardware will keep rising as more Chinese hyperscalers shift to local suppliers.
This shift reflects a broader trend inside China’s technology sector, where companies are adapting their long-term strategies around secure, compliant and locally sourced components.
Pressure grows as chip supply tightens
China’s AI sector is facing sustained shortages in advanced chips. Alibaba stated this month that components required for building data centres will remain a major bottleneck for the next few years.
Tencent reported that its capital spending for next year will be lower than expected, not because of weaker demand but because chips are unavailable at the scale needed.
The shortage is shaped by both global and domestic factors. High international demand has created bottlenecks across the semiconductor supply chain, while US rules prevent Nvidia from selling its top-end AI processors to China.
Beijing has also advised companies to avoid relying on the H20, a lower tier Nvidia chip designed for the Chinese market but cleared for export.
Chinese firms have responded by relying on stockpiled components and making their AI models more efficient so they can operate with fewer chips.
However, manufacturing challenges persist. China’s largest chipmaker, SMIC, is unable to match the scale or technology of Taiwan Semiconductor Manufacturing Co, which limits the country’s ability to produce advanced processors domestically.
Competition rises as tech firms build their own chips
Baidu is not alone in developing in-house semiconductors.
Alibaba is working on its next-generation AI chip, while other companies are adjusting their infrastructure plans to manage supply constraints.
Analyst estimates suggest Kunlunxin could achieve significant valuation growth, supported by strong revenue forecasts for the coming years.
The combination of restricted imports, strong domestic demand, and long-term self-reliance goals has created a larger market for companies capable of building competitive AI chips.
Baidu’s strategy places it in a strong position to become a critical supplier across cloud computing, telecom networks, and enterprise AI workloads.
The post Baidu pushes deeper into AI chips as China races to replace Nvidia appeared first on Invezz
