Huawei Technologies says it will launch two new artificial-intelligence chips in 2027 as it steps up efforts to build a domestic alternative to Nvidia and expand China’s ability to train and run advanced AI systems without relying on US technology.

Rotating chairman David Wang said at Huawei Connect in Shanghai on Thursday that the company plans to release the 960DT in the first quarter of 2027, followed by the Ascend 960PR in the third quarter. Huawei also highlighted its UnifiedBus interconnect technology, which is designed to connect large numbers of processors so they can work together as a single computing system.

The strategy reflects Huawei’s growing focus on system-level computing rather than trying to beat Nvidia solely on the performance of an individual chip. The company has already shipped more than 1,000 AI systems using its supernode architecture to more than 370 customers, and says its networking technology can eventually support clusters containing extremely large numbers of processors.

That approach has become increasingly important because Huawei continues to operate under US restrictions that limit its access to advanced semiconductor manufacturing equipment and high-end memory. Chinese chipmakers have also been hit by a shortage of high-bandwidth memory, a key component in AI accelerators.

Huawei is nevertheless gaining ground inside China. Demand for its Ascend processors has risen as major Chinese technology companies look for alternatives to Nvidia, particularly after Beijing encouraged greater use of domestically produced hardware. Earlier this year, companies including ByteDance, Tencent and Alibaba were reported to be seeking additional Huawei chip supplies after the launch of DeepSeek’s V4 model. Huawei’s Ascend 950PR was reported to outperform Nvidia’s older China-focused H20 in some workloads, although it still trailed Nvidia’s more powerful H200.

Nvidia remains the global leader in AI computing and retains a major advantage through its CUDA software ecosystem, which has become deeply embedded in AI development around the world. Huawei’s challenge is therefore not simply to produce faster silicon, but to convince developers and large cloud customers that its hardware, networking and software can work together reliably at scale.

The 2027 launches also represent an acceleration of Huawei’s previously announced roadmap. Last year, the company said the Ascend 960 would arrive in the fourth quarter of 2027 and promised roughly double the computing power, memory capacity and interconnect performance of the Ascend 950 generation. The newly announced 960DT and 960PR schedule brings multiple variants into the year and suggests Huawei is broadening the line more aggressively than originally indicated.

The announcement comes as competition between the United States and China over AI hardware continues to intensify. Washington has restricted China’s access to the most advanced Nvidia processors, although limited shipments of the H200 have recently been allowed under licence. At the same time, Chinese companies are investing heavily in domestic chips to reduce their dependence on US suppliers.

Huawei is still some distance from displacing Nvidia globally. But its 2027 roadmap shows that the competition is increasingly moving beyond individual processors and towards complete AI computing systems, where China is betting that scale and integration can compensate for constraints at the cutting edge of chip manufacturing.