Deepseek’s models drive the competitiveness of Chinese chip manufacturers against US rivals

The emergence of artificial intelligence (AI) models of Deepseek is seen that provides some manufacturers of Chinese chips such as Huawei for a better opportunity to compete in the domestic market against more powerful US processors.

It is seen that some manufacturers of Chinese chips, such as Huawei, have a greater possibility of competing in the local market against more powerful American processors thanks to the development of artificial intelligence models (AI) of Deepseek. (AP) (AP)

Huawei and their Chinese partners have fought for years to coincide with NVIDIA in the construction of high -end chips that could compete with the products of the US company. UU. For training models, a process in which data is They feed algorithms to help them learn to make precise decisions.

However, Depseek models, which focus on “inference”, or when an AI model produces conclusions, optimizes computational efficiency instead of depending solely on unprocessed processing power.

Read too: Avoid the dubsant application of Deepseek and the Proton VPN censorship battle

That is a reason why the model is expected to close the gap between what the Chinese manufacturing processors and their most powerful American counterparts can do, they say analysts.

Huawei and other Chinese chips manufacturers such as Hygon, Tencent backed by threads, Tsingmicro and Moore have issued in the latest statements that claim that the products will admit Deepseek models, although few details have been published.

Huawei declined to comment. Moore Threads, Hygon Enflame and Tsingmicro did not respond to Reuters consultations for more comments.

Industry executives now predict that the Open Source Nature of Deepseek and its low rates could boost the adoption of AI and the development of real -life applications for technology, helping Chinese companies overcome the export edges Americans in their most powerful chips.

Even before Depseek arrived at the headlines this year, customers such as customers, such as bytete, such as less intensive “inference tasks, such as” inference “tasks, such as” inference tasks “, as the realization tasks, such as realization tasks, such as” inference “tasks, such as less intensive” inference “tasks. Through chatbots.

In China, dozens of companies, from car manufacturers to telecommunications suppliers, have announced plans to integrate Depseek models with their products and operations.

“This development is very aligned with the ability of Chinese chipset suppliers of AI,” said Lian Jye, Chief Analyst of the OMDIA Technological Research firm.

“Chinese the Chinese chips fight to compete with the NVIDIA GPU (graphics processing unit) in AI training, but the inference workloads of AI are much more indulgent and require a much more local and specific understanding of The industry, “he said.

Nvidia still dominates

However, Bernstein analyst Lin Qingyuan said that while Chinese chips were competitive in the cost for inference, this was limited to the Chinese market since Nvidia chips were still better even for inference tasks.

Although US export restrictions prohibit the most advanced NVIDIA AI training chips to enter China, the company can still sell less powerful training chips than Chinese customers can use for inference tasks.

Nvidia published a blog post on Thursday on how inference time increased as a new scale law and argued that its chips will be necessary to make Veteran and other more useful “reasoning” models.

In addition to the computer power, the CUDA of NVIDIA, a parallel computing platform that allows software developers to use the NVIDIA GPU for general use computing, not only AI or graphics, it has become a crucial component of Your domain.

Previously, many Chinese AI Chips companies did not directly challenge Nvidia by asking users to leave CUDA, but instead, they affirmed that their chips were compatible with CUDA.

Also read: IITM launches joint research projects to revolutionize climatic predictions using AI

Huawei has been the most aggressive in his efforts to separate from Nvidia by offering a CUDA equivalent called Compute Architecture for Neural Networks (CANN), but experts said they faced obstacles to persuade developers to leave CUDA.

“The software performance of Chinese companies of artificial intelligence chips at this stage is also .