What Drives Nvidia’s AI Dominance and How Huawei’s Ascend Chips Compete

Q1: How to View Nvidia’s Development

Nvidia, founded in 1993, reshaped the computing industry through forward‑looking technology and ecosystem building. After its 1999 NASDAQ IPO, the company launched the revolutionary CUDA parallel‑computing architecture in 2006, extending GPUs from graphics rendering to high‑performance computing and laying the groundwork for the AI boom.

In 2016 Nvidia introduced the world’s first deep‑learning supercomputer DGX‑1, establishing a benchmark for AI infrastructure. The company now offers a four‑track product matrix covering data‑center, gaming, professional visualization, and autonomous driving, with flagship GPUs such as the Turing architecture and the Hopper‑based H100 accelerator continuing to push industry innovation.

Q2: What Is the Status of Huawei’s Ascend Series?

Huawei’s Ascend chips are deeply integrated into China’s “new infrastructure” and “East‑Data‑West‑Compute” strategies, emphasizing domestic, controllable AI compute. The first Ascend 310 edge AI processor and the all‑scenario AI framework MindSpore were released in 2018, marking a major breakthrough in China’s AI foundation infrastructure.

Through continuous iteration, the Ascend portfolio now includes the Ascend 910 training processor, Atlas compute clusters, and the MindSpore 2.0 framework. The Da Vinci architecture delivers a 30% higher FP16 compute density than the industry average, while the heterogeneous CANN 6.0 stack supports end‑to‑end AI workflows from model development (MindStudio) to deployment (ModelArts).

Q3: What Advantages Does the Ascend 310 Chip Offer?

The Ascend 310 is a system‑level AI chip (SoC) optimized for multimodal data processing. Its Da Vinci architecture features a three‑dimensional cube‑core design that enables instruction‑level parallelism. Each chip integrates two AI‑Core clusters, each containing 32 AI compute units (CUBE Core), supporting mixed‑precision FP16/INT8 operations with a peak performance of 16 TOPS@FP16 while consuming only 8 W, making it ideal for edge devices.

Q4: Recent Highlights of Huawei Ascend (2024‑2025)

2024: Release of CANN 8.0 and MindSpore 2.4, alongside nationwide deployment of upgraded compute centers.

2025 (planned): Joint launch with Silicon Flow and Huawei Cloud of the Ascend‑based DeepSeek R1/V3 inference service.

Q5: Which Companies May Benefit from Ascend’s Growth?

Hardware providers and AI‑related firms stand to gain from the domestic AI compute upgrade. Notable beneficiaries include iFlytek, SMIC, Tuorui, Rtong, Huafeng Technology, Guangdian Yuntong, Digital China, Bowei Alloy, Sichuan Changhong, and Shaanxi Huada, among others.