- Ascend 910D: the chip of IA most advanced of Huawei, designed to compete with the Nvidia H100.
- US sanctions on Nvidia boost Huawei's position as an alternative in the Chinese market.
- The 910D GPU is currently in testing, with samples expected by the end of May.
- Huawei seeks to lead the AI industry in both inference and model training.
In the last months, Huawei has redoubled its efforts to position itself as a key player in the food processor sector Artificial Intelligence in China. The international context, especially the US restrictions Nvidia's technology exports have caused a real shake-up in the market. Giants like Tencent, Alibaba, and ByteDance, traditional Nvidia customers, are now looking for alternative solutions to maintain their leadership in innovation and AI processing.
Against this backdrop, Huawei begins a new stage in the technological race presenting the Ascend 910D, a processor destined to break US dominance thanks to its technical capacity and the favorable circumstances resulting from international sanctions. The company, previously focused primarily on chips for inference tasks, now aims to compete head-to-head in model training as well, a significant leap forward from its previous strategy.
Ascend 910D: The chip Huawei hopes to use to lead AI
The new GPU Huawei Ascend 910D is in the validation process, and the first samples are expected to reach Chinese companies for evaluation by the end of May. According to media outlets such as The Wall Street Journal and Reuters, the chip has been specifically designed to outperform the Nvidia H100, the current benchmark in raw power and energy efficiency in the AI sector.
The 910D joins the Ascend family, which already included models such as the 910B and 910C, used primarily for inference tasks by large technology companies in the country. The new model introduces advanced packaging technologies, allowing more silicon components to be incorporated into less space and with greater processing capabilities. Although it still consumes more energy than its American rival, Huawei is confident that efficiency will continue to improve as the development and testing phases progress.
US sanctions on the export of advanced processors have left Nvidia's traditional Chinese customers with few viable options. Given the export ban on the H20 GPU and the previous restriction on the H100, the demand for high-performance alternatives In the local market, it has skyrocketed. With its Ascend 910D, Huawei aims to become that national option capable of reducing technological dependence on the West.
A strategy to conquer AI inference and training
Until recently, Huawei focused on developing chips optimized for the AI inference, that is, to run models and generate responses to users or processes. However, with the 910D, the company also aims to cover the needs of training, and of models, a function that requires a even greater computing power and is usually in the hands of the most advanced processors in the world.
The ability to combine both uses in a single chip is what distinguishes Huawei's new developments, which thus seeks equate–or even overcome– the performance of Nvidia's leading solutions in the global market. According to experts and sources close to the company, the goal is ambitious: lead global innovation in AI processors and ensure China's technological autonomy in this strategic field.
The challenge is enormous, as large-scale manufacturing problems persist due to the lack of access to cutting-edge technologies from companies such as TSMCHuawei relies on the capabilities of its local partner SMIC, although this company also works with restrictions on access to cutting-edge machinery. Despite this, the Chinese technology company already plans to ship more than 800.000 Ascend 910B and 910C chips to local customers this year, consolidating its presence while the 910D makes its way into the market.
Ascend 910D and the new geopolitics of AI chips
The development and deployment of the Ascend 910D It occurs in a context marked by geopolitical tensions between China and the United States, where the fight for control of artificial intelligence is considered strategic. The Chinese government sees these advances as a key tool for achieving technological autonomy. and avoid limitations imposed from abroad.
Industry experts suggest that the arrival of this GPU could shift the international balance, allowing the Chinese tech ecosystem to have its own high-end solution. In parallel, the company is developing other infrastructures, such as CloudMatrix 384, designed to optimize the collaborative work of multiple Ascend chips and thus maximize performance in distributed AI tasks.
This launch reflects a strategy focused on consolidating China's position in the AI chip market and reducing its dependence on foreign technologies, strengthening its technological autonomy and ability to compete internationally.
Passionate writer about the world of bytes and technology in general. I love sharing my knowledge through writing, and that's what I'll do on this blog, show you all the most interesting things about gadgets, software, hardware, tech trends, and more. My goal is to help you navigate the digital world in a simple and entertaining way.