Huawei Technologies is getting closer to a launch of two AI chips in 2027: the first generation 960DT in the first quarter of 2027 and the Ascend 960PR in the third quarter. The announcements underscore Huawei's growing embrace of AI at a time of shortage of advanced foreign semiconductors.
The planned chips are part of Huawei’s Ascend AI computing strategy, which has become increasingly important as Chinese technology companies look for locally developed alternatives to high-end processors from US companies. The 960DT is expected in the first quarter of 2027 and the Ascend 960PR in the third quarter.
Huawei’s new roadmap demonstrates that the company is going to continue to develop AI hardware over time, rather than the same chip generation once and for all. As computing capacity increases, so does the speed of development and deployment of generative AI applications.
In addition to individual processors, Huawei is also developing an interconnection technology called UnifiedBus. It is designed for large numbers of AI chips to communicate and function together as a unified computing system. Huawei claims to have developed 11 semiconductors with the technology and has set up large AI superclusters that can be home to one million AI processors.
Such systems are important because more advanced AI models demand huge computing power, so the communication between processors is as important as the performance of individual chips.
Huawei has also been expanding the use of smaller, interconnected AI systems called supernodes. More than 1,000 supernodes have now been delivered to more than 370 customers, the company said.
Each supernode integrates AI chips to perform computing tasks in a coordinated manner. This is representative of a larger trend in AI infrastructure: companies are using large numbers of processors and high-speed networks to handle more demanding workloads rather than just one processor.
The development of Huawei's AI chip business has taken place in the context of US restrictions on China's access to advanced semiconductors and chipmaking technologies. Those restrictions have complicated Chinese companies’ access to some of the most advanced AI processors available internationally.
At the same time, Chinese technology companies have become increasingly focused on domestic alternatives. Huawei is thus a key player in the construction of China’s semiconductor and AI ecosystem by being independent.
The launch of these chips for the 2027 chips will therefore be closely watched by the global semiconductor industry. Nvidia is still a key player in AI computing technology and has a large software ecosystem in place in AI computing, not only with processors but also with a software ecosystem and developers. Huawei's strategy extends more than just to developing individual chips from a chip-to-chip basis and includes processors, interconnection technology and large-scale computing systems.
How well its new products compete with Nvidia in terms of performance, energy efficiency, software compatibility and large-scale deployment will depend on the technical features and how well their customers will use them for real-world applications. In the short term, Huawei’s 2027 roadmap for AI hardware investment is likely to continue in the future because China is pressing ahead with the development of AI hardware for its domestic computing needs.
