Will Huawei’s New AI Chip, Ascend 910D, Rival Nvidia?

With countries seeking technological independence as export controls reshape global supply chains, China's access to the high-performance processors required for training advanced AI systems has been impacted.

In response, Huawei has developed a new AI processor that aims to match capabilities of restricted US hardware.

The company hopes this latest iteration of its Ascend AI processors will provide performance comparable to Nvidia's H100 graphics processing unit (GPU), according to Reuters.

Huawei’s Ascend processors targeting computing independence

This development reportedly follows Huawei’s plans to begin mass shipments of its previous 910C AI chip to Chinese customers as early as next month, suggesting the company is pursuing a multi-generation strategy to address different segments of the AI computing market.

Huawei and other Chinese semiconductor manufacturers have struggled for years to match the performance of Nvidia, as its architecture sets industry standards for the specialised processing required to train models through a process where extensive data feeds algorithms to help them learn decision-making.

However, Washington has increasingly restricted China's access to advanced semiconductor technology as part of efforts to limit China's technological development, particularly advances with potential military applications.

Additionally, the US government barred the sale of Nvidia's H100 chip to Chinese customers in 2022 before the product officially launched.

US authorities have since expanded restrictions to include Nvidia's newest B200 chip, further limiting Chinese companies' access to hardware necessary for training large language models and other advanced AI systems.

Chinese technology sector adapting to export controls

These export controls have created significant obstacles for Chinese AI researchers and companies attempting to train sophisticated models, which require substantial computing resources.

The computational requirements for training state-of-the-art AI systems have increased exponentially in recent years.

Now, the practical gap between Huawei's chips and Nvidia alternatives extends beyond raw computational power.

Nvidia's CUDA software platform, which provides tools for developers to use GPUs for general-purpose computing, has become the industry standard with widespread adoption. Meanwhile, Huawei's Ascend chips use the company's Compute Architecture for Neural Networks (CANN), which lacks the mature ecosystem and developer familiarity of CUDA.

For Huawei, success in the AI chip sector would mark a significant milestone in the company's efforts to diversify its business following US sanctions that severely impacted its smartphone division.

Huawei’s AI ambitions amid geopolitical tensions

Huawei’s push into AI processors is part of a broader strategy to reduce reliance on foreign technology amid escalating US-China tensions.

Since being placed on the US Entity List in 2019, which restricted its access to American components and software, Huawei has increasingly focused on enterprise technology, cloud computing and semiconductor development to sustain growth.

The Ascend 910D chip is an important component of this pivot, targeting the lucrative and fast-growing AI computing market.

China’s government has also made semiconductor self-sufficiency a national priority, investing billions in research and development to support domestic chipmakers.

However, the challenge is not only technical but also commercial.

Nvidia’s GPUs benefit from a vast ecosystem of AI researchers, developers and cloud providers who rely on Nvidia’s CUDA platform for efficient model training and deployment.

Whereas Huawei’s CANN platform, while promising, is still developing the same level of software compatibility and developer tools, which could slow adoption outside China.

Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. 

Discover all our upcoming events and secure your tickets today.

Also sign up to our free weekly newsletter for the latest insights and stories straight into your inbox.

AI Magazine is a BizClik brand