China’s AI chipmakers get competitive boost from DeepSeek

#Chinas #chipmakers #competitive #boost #DeepSeek

The appearance of Dipic’s artificial intelligence (AI) models is offering a strong step in the domestic market like Chinese chipmakers like Huawei, allowing it to better compete against powerful American processors.

For years, Huawei and other Chinese firms have faced challenges to meet NVIDIA’s high-end chips, which dominates the AI ​​training sector-where the algorithmic decision-making to improve decision-making Goes

However, the Dippic models prefer “inadays” – the phase where the AI ​​concludes.

This is the reason why the model is expected to partially close the gap between Chinese -made AI processors and their more powerful US counterparts, analysts say.

Huawei, and other Chinese AI chipmakers such as Hygon, Tensent -backed Enfolm, Sisingmro and Moore Threads have issued statements in recent weeks claiming that the product will support the Deep SEC model, though some details Have been released.

Huawei refused to comment. Moore Threads, Hygon Enfall and Singmro did not answer Writers’ questions to comment further.

Industry executives are now predicting that the open source nature of DiPsk and its low fees can increase the development of real -life applications for the adoption of AI and technology, which makes Chinese firms their powerful chips But preventing US exports can help overcome.

This year, before DiPsic made headlines, products like Huawei -climbing 910B were seen by users like Betnsus, such as low -computation is better suitable for “in conference” works, included training AI models in post -training AI model There are, like they work or work, like they work. Chat through boats.

In China, dozens of companies, from car makers to telecom providers, have announced plans to connect Deep Sak models with their products and operations.

“This development is much more associated with the capacity of Chinese AI chip set vendors,” said Lean Jae SU, a chief analyst at the tech research firm Omdia.

“The Chinese AI Chip is struggling to compete with the GPU (graphics processing unit) in the AI ​​training, but the work burden about AI is too much forgiven And it requires a lot of local and specific industry understanding. “

Nvidia still gains dominance

However, Burnstein’s analyst Lin Genghian said that while Chinese AI chips were competitive for infreening, it was limited to the Chinese market because NVIDIA chips were still better for better tasks.

Although US export sanctions ban NVIDIA’s latest AI training chips are prohibited from entering China, the company is still allowed to sell less powerful training chips that can be used for Chinese -grown individualism.

NVIDIA published a blog post on Thursday how to estimate the time as a new scaling law and argued that its chips are necessary to make DiPsic and other “reasoning” models more useful. Will

In addition to computing power, NVIDIA’s CUDA, a parallel computing platform that allows software developers to use NVIDIA GPUs for general purpose computing, not only became an important component of AI or graphics, its domination. Is

Earlier, many Chinese AI chip companies asked users to quit coda, but instead, their chips did not challenge NVIDIA directly to be compatible with Koda.

Huawei has been the most aggressive in efforts to separate from Nvidia by offering a CUDA called CUDA for Neural Networks (CAN), but experts say he faces obstacles to the developers to abandon CUDA Have to do

“At this stage, Chinese AI chip firms also lack software performances,” said Omdia’s SU. Koda has a full -fledged library and a diverse range of software capabilities, which requires significant long -term investment. “