This site is fictional demo content. It is not real news or affiliated with any real organization. Do not treat it as fact or professional advice.

NextPaper Logo
nextpaper.ccCYBER CHRONICLE
RSS
EN中文

Full article

FULL TEXT

View this issue
Deep diveAI

Rise of Chinese AI Chips: The New Landscape of Domestication in 2028

Against the backdrop of tightening U.S. export controls, Chinese AI chips achieved key breakthroughs in 2028. Huawei Ascend, Cambricon, and Inspur's latest products approach international advanced levels, with ecosystem building becoming the key to victory.

Content

In 2028, China's AI chip industry delivered a surprisingly positive report card.

Performance Approaching the First Tier

Huawei Ascend 910C's actual performance in mainstream LLM training scenarios has reached 78%-85% of NVIDIA H100, significantly improved from 65% in 2027. Cambricon's latest generation Suiyuan-3 chip has surpassed A100's energy efficiency ratio in inference scenarios by approximately 15%. Inspur Technology achieved an overperformance against H100 in specific AI workloads through its unique sparse computing architecture.

These numbers represent comprehensive breakthroughs achieved by Chinese AI chips through architectural innovation, packaging improvements, and software ecosystem optimization, despite advanced process constraints.

Ecosystem Pain and Ecosystem Solutions

However, hardware performance catch-up is only part of the story. The CUDA ecosystem moat remains high—over 85% of global AI training frameworks and toolchains are built on CUDA, and developers are accustomed to CUDA's programming models and debugging tools. The switching cost is far higher than expected.

The solution lies in "software-hardware integration": Huawei's continuous investment in the CANN ecosystem has initially established a migration toolchain covering mainstream frameworks; an industry alliance formed by multiple AI startups is promoting unified runtime standards and model formats to reduce developer migration friction.

Commercial Deployment Accelerates

In Q1 2028, Chinese AI chips' share among Chinese cloud service providers increased from approximately 20% in 2026 to approximately 41%. A procurement director at a major cloud provider stated: "At equivalent performance, Chinese chips' price advantages and localized service response speed already offset the hidden costs from ecosystem gaps."

Demand for localized LLM deployment has become a new growth engine for Chinese chips. In 2028, an increasing number of government and enterprise customers are choosing private LLM deployment, which aligns with Chinese chips' advantages in data security and cost.

Boundary

This is fictional content for entertainment only.