The deal underscores Beijing’s strategy of building a stronger homegrown chip ecosystem to rival Nvidia, whose GPUs and CUDA software platform currently dominate global AI development.
Kunlunxin said its AI chips, which are compatible with Nvidia’s CUDA framework, will be supplied to H3C Technologies and telecoms equipment giant ZTE—two leading suppliers to China Mobile. The move is seen as a crucial step in ensuring developers can more easily transition from Nvidia’s ecosystem to Chinese-made alternatives without incurring prohibitive costs.
According to corporate registry data from Qichacha, Baidu holds a 59% stake in Kunlunxin. The tech giant has been aggressively scaling its chip operations, unveiling in April a cluster of 30,000 third-generation P800 Kunlun chips designed to power large-scale AI training. The chips, Baidu claims, can support models comparable to those developed by Chinese AI startup DeepSeek.
China Mobile’s procurement strategy also included orders for hardware compatible with Huawei’s Compute Architecture for Neural Networks (CANN)—Huawei’s in-house equivalent to CUDA. Tender results revealed that a Huawei-linked entity won a portion of the contracts, marking another milestone in China’s bid to diversify away from Nvidia.
While Huawei’s Ascend chips remain prominent in the market, analysts note that Baidu’s Kunlun processors enjoy stronger CUDA compatibility, which reduces migration hurdles for developers accustomed to Nvidia’s platform. This technical edge may prove decisive as a dozen Chinese-made alternatives compete for market share in a market long dominated by Nvidia GPUs.
The orders come at a pivotal moment as Chinese authorities intensify efforts to build a self-sustaining semiconductor industry amid ongoing U.S. export restrictions on advanced chips and chipmaking tools. For Beijing, fostering viable domestic GPU options is not only about technological independence but also about ensuring that China’s fast-growing AI sector can scale without disruption.