China-based artificial intelligence initiative DeepSeek has achieved an important milestone in the industry by choosing domestic Huawei Ascend chips instead of Nvidia or AMD in the training processes of the new generation DeepSeek V4 Pro model it developed. In the project carried out in cooperation with Shenzhen-based institutions, the post-training process of the model was successfully completed using approximately 1000 Ascend 910C chips. This development is considered as a strategic step taken by the Chinese artificial intelligence industry, which is struggling with US-based advanced chip supply restrictions, to reduce foreign dependency. Previously, the DeepSeek V3 model was trained on 2048 Nvidia H800 chips.

The use of Huawei Ascend chips creates a critical threshold in China’s aim to establish a self-sufficient artificial intelligence ecosystem.

Domestic Hardware Stands Out in Artificial Intelligence Training

The development process of artificial intelligence models generally consists of two basic phases: data processing and the ability of the model to understand instructions. Previously, Chinese companies were limited to Nvidia’s high-performance hardware, especially in processes that required complex model training.

However, this step taken with DeepSeek V4 Pro showed that Huawei’s Ascend architecture has the capacity to successfully train modern large language models (LLM).

The success of the project was made possible by the collective effort of academic and industrial stakeholders such as Shenzhen Loop Area, Harbin Institute of Technology and Shenzhen Greater Knowledge Institute.

Dependency in the Chip Market Begins to Reduce

In the past, models like DeepSeek V3 have been trained based on large-scale Nvidia H800 systems. However, current export restrictions have made access to such advanced hardware difficult. This breakthrough by DeepSeek stands out as not only a test but also an operational solution. The performance provided by Huawei Ascend 910C chips gives a strong signal that Chinese developers can now get rid of the forced dependence on Nvidia or AMD.

Offering a high cost advantage, DeepSeek plans to increase its competitiveness with domestic hardware.

New Generation Models Continue to Attract Investment

DeepSeek’s rapid rise in the field of artificial intelligence also attracts the attention of investors. It is stated that the company is preparing for a huge financing type worth approximately 7.4 billion dollars. DeepSeek, which aims to offer lower-cost and higher-performance solutions against ChatGPT and other global Competitors, has the potential to create a revolution in areas such as text production and code writing. On the hardware side, this strategic partnership established with Huawei ensures the long-term sustainability of the company.

How do you think Chinese companies’ transition to domestic chips will affect global artificial intelligence competition? Share your intentions with us in the comments.