Huawei Ascend 910C vs. NVIDIA H100: A big leap towards AI Hardware revolution...

Huawei is looking to take on semiconductor behemoth Nvidia with a new advanced AI chip.

Huawei is making progress developing its latest Ascend AI GPU, the Ascend 910D. The Company is hoping that its chip will rival Nvidia’s H100 series, which is popular for training AI models.

Huawei's 910C, a graphics processing unit (GPU), represents an architectural evolution rather than a technological breakthrough, according to one of the two people and a third source familiar with its design.

It achieves performance comparable to Nvidia's H100 chip by combining two 910B processors into a single package through advanced integration techniques, they said.

That means it has double the computing power and memory capacity of the 910B and it also has incremental improvements, including enhanced support for diverse AI workload data, they added.

 As the global demand for AI accelerates, Huawei's Ascend 910C emerges as a formidable competitor to NVIDIA's H100, especially within China's rapidly evolving tech landscape.

⚙️Performance & Efficiency

While the Ascend 910C's raw performance metrics trail the H100's, especially in FP16 and INT8 operations, it offers competitive BF16 performance at 780 TFLOPS. Notably, the 910C operates at approximately 310W, making it more power-efficient compared to the H100's 700W consumption.

In practical applications, such as inference tasks for large language models like DeepSeek R1, the Ascend 910C delivers about 60% of the H100's performance.

🌐 Strategic Positioning

Huawei's Ascend 910C is more than just a chip; it's a strategic move towards technological self-reliance. With U.S. export restrictions limiting access to NVIDIA's advanced GPU s, the 910C provides a domestically produced alternative for China's AI industry.

Collaborations with tech giants like Baidu, ByteDance, and Tencent have facilitated the integration of Ascend chips into cloud services and data centres.

📈 Market Outlook

Huawei plans to ship over 100,000 units of the Ascend 910C by the end of 2025, reflecting aggressive growth intentions. An initial order of 70,000 chips, valued at $2 billion, suggests a competitive pricing strategy.

🧠 Software Ecosystem

While NVIDIA's CUDA platform remains a dominant force in AI development, Huawei is investing in its CANN (Compute Architecture for Neural Networks) and MindSpore frameworks to build a robust software ecosystem. These tools aim to provide developers with alternatives for AI model development and deployment.

🏁 Conclusion

Huawei's Ascend 910C represents a significant step towards diversifying the AI hardware landscape. While it may not yet surpass NVIDIA's H100 in raw performance, its strategic value, energy efficiency, and integration into China's AI ecosystem make it a noteworthy contender in the global AI race.