All Topics
All Topics
Technology
Technology
AI
AI
Business
Business
Entertainment
Entertainment
News
News
Programming
Programming
Security
Security
Science
Science
Design
Design
Environment
Environment
Finance
Finance
Crypto
Crypto
Politics
Politics
Sports
Sports
Education
Education
Gaming
Gaming
Art
Art
Music
Music
Health
Health
Books
Books
Food
Food
Travel
Travel
Personal
Personal
Bluesky
Twitter

China trains trillion-parameter AI model on domestic chips, challenging Nvidia's training monopoly

By

Poe Zhao

1d ago· 4 min readenInsight

Summary

The article examines whether China can train frontier AI models without Nvidia chips, using Meituan's LongCat-2.0 as key evidence. The trillion-parameter model was pre-trained from scratch on a 50,000-chip cluster powered entirely by domestic processors (Huawei Ascend). This challenges the long-held assumption that Chinese chips could only handle inference, not training. The piece maps the capability ladder and efficiency gap between Chinese domestic chips and Nvidia's offerings, analyzing implications for investors and policymakers.

Source

Twitter / XChina trains trillion-parameter AI model on domestic chips, challenging Nvidia's training monopolyhellochinatech.com

Key quotes

· 3 pulled
For 2 years, the working assumption among investors and policymakers has been clear: Chinese AI chips could handle inference... Training, the far more demanding process of building one, still required Nvidia.
In late June, Meituan, China's largest on-demand services platform, released LongCat-2.0.
The company says the trillion-parameter model was pre-trained from scratch on a 50,000-chip cluster powered entirely by domestically produced processors.
Snippet from the RSS feed
Meituan's LongCat-2.0 trained on 50,000 domestic chips. Huawei's Ascend enters frontier training. We map the capability ladder and the efficiency gap.

You might also wanna read

Comments

Sign in to join the conversation.

No comments yet. Be the first.