Huawei Struggles to Break NVIDIA's AI Chip Dominance in China Despite Homegrown Push

Huawei's Struggle to Gain Traction in China's AI Chip Market

Huawei Technologies, China's tech giant, is facing significant challenges in its attempt to break NVIDIA's stronghold on the AI chip market in China. Despite rising interest in homegrown solutions, major Chinese tech companies are hesitant to adopt Huawei's AI chips, according to recent reports 1

1

2

2

.

Technical Challenges and Ecosystem Lock-in

Huawei's Ascend 910C chip, which combines two older 910B chips to deliver 800 TFLOP/s of computing power at FP16, is considered comparable to NVIDIA's H100 GPU 1

1

. However, the chip has encountered several technical issues:

1. Overheating problems affecting reliability perceptions
2. Limitations in Huawei's software stack, CANN (Compute Architecture for Neural Networks), compared to NVIDIA's industry-standard CUDA platform
3. Lack of native support for memory-optimal calculation formats like FP8

These technical shortcomings have made it difficult for Huawei to convince major Chinese tech companies like ByteDance, Alibaba, and Tencent to place significant orders for its AI chips 2

2

.

NVIDIA's Ecosystem Advantage

NVIDIA's dominance in the AI chip market is largely attributed to its well-established ecosystem:

1. Many Chinese tech companies have heavily invested in NVIDIA's CUDA ecosystem
2. Switching to Huawei's chips would require significant time and resources to redesign data center infrastructure, software, and engineering workflows
3. NVIDIA's existing inventory among Chinese tech companies reduces the urgency to switch

Geopolitical Factors and Export Controls

The situation is further complicated by geopolitical tensions and US export controls:

1. The US Department of Commerce issued guidance in May, potentially deeming companies using Huawei's chips without authorization in violation of US export controls 1

   1

2. This has particularly affected Chinese entities with significant overseas presence
3. At least one Chinese data center firm has canceled a tentative order for Huawei hardware due to these concerns 2

   2

Huawei's Response and Future Prospects

Huawei is not giving up on its AI chip ambitions:

1. The company has unveiled CloudMatrix 384, bundling up to 384 Ascend chips to provide comparable computing power to NVIDIA's supercomputers 1

   1

2. Huawei has created a translation tool to induce artificial compatibility with FP8, although this solution remains suboptimal
3. The company has secured sales to state-owned enterprises and local governments in China 2

   2

The Role of Open-Source Developers

The future of Huawei's AI chip ecosystem may depend on the support of open-source developers:

1. Companies like DeepSeek could play a pivotal role in shaping Huawei's trajectory
2. If DeepSeek adopts Huawei's chips and software extensively, it could attract many open-source developers to build on Huawei's ecosystem 2

   2

As the AI chip race continues, Huawei faces an uphill battle to break NVIDIA's dominance in China's tech sector. The company's success will depend on addressing technical challenges, building a robust ecosystem, and navigating complex geopolitical factors.