Huawei's Ascend 910C Challenges Nvidia's AI Dominance with 60% H100 Inference Performance

4 Sources

Share

Huawei's Ascend 910C AI chip, developed under US sanctions, achieves 60% of Nvidia H100's inference performance. This breakthrough could reduce China's reliance on US tech and disrupt the global AI hardware market.

News article

Huawei's Ascend 910C: A Challenger to Nvidia's AI Dominance

In a significant development for the AI hardware industry, Huawei's Ascend 910C chip has demonstrated impressive performance capabilities, achieving 60% of the inference performance of Nvidia's flagship H100 GPU. This breakthrough, revealed by AI research firm DeepSeek, marks a potential shift in the global AI chip market and highlights China's progress in reducing its dependence on US technology

1

.

Technical Specifications and Manufacturing

The Ascend 910C, built on SMIC's 7nm N+2 process, boasts 53 billion transistors and utilizes chiplet packaging. Despite being manufactured without access to TSMC's cutting-edge technology due to US sanctions, the chip has managed to punch above its weight, particularly in inference tasks

3

.

Performance and Optimization

DeepSeek's research indicates that the Ascend 910C's performance exceeded expectations, especially in inference workloads. The chip's efficiency can be further improved through manual optimizations of CUNN kernels. DeepSeek's native support for Ascend processors and its PyTorch repository facilitates seamless CUDA-to-CUNN conversion, easing the integration of Huawei's hardware into AI workflows

3

.

Market Implications

The emergence of the Ascend 910C could potentially disrupt Nvidia's stronghold on the AI chip market. As a cost-effective alternative for inference tasks, it may appeal to companies looking to avoid Nvidia's premium pricing. This development has led to speculation about potential market sell-offs and negative retail reactions, particularly if Huawei times the release of the Ascend 910C to coincide with Nvidia's GTC 2025 keynote

2

.

Challenges and Limitations

While the Ascend 910C shows promise in inference tasks, it still faces challenges in AI training capabilities. Long-term training reliability remains a critical weakness for Chinese processors, largely due to the deep integration of Nvidia's hardware and software ecosystem developed over two decades

3

.

Future Prospects

As AI models increasingly converge on Transformer architectures, the importance of Nvidia's software ecosystem may decline. This shift, coupled with DeepSeek's optimization tools and expertise, could further reduce dependency on Nvidia, offering AI companies a more cost-effective alternative, particularly for inference tasks

4

.

Geopolitical Implications

The development of the Ascend 910C underscores China's progress in domestic chip production despite US sanctions. This advancement may fuel further innovation in China's AI industry, potentially altering the global AI landscape and challenging US technological dominance

2

.

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2025 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo