Alibaba uses its own chips to train next-generation Qwen model
Alibaba is using chips from its semiconductor subsidiary T-Head to train its upcoming Qwen model, CEO Eddie Wu disclosed on Thursday during the company’s quarterly earnings call. This decision integrates T-Head’s hardware into both the training and inference stages of Alibaba's AI operations.
Alibaba CEO Eddie Wu stated on Thursday that chips developed by its semiconductor subsidiary, T-Head, are now being used to train its next Qwen model. This announcement, made during the company’s quarterly earnings call, confirms T-Head’s involvement in a critical phase of AI development. Additionally, Alibaba’s Bailian model platform, known internationally as Model Studio, is providing inference services, running AI models to answer user requests, using Chinese-made chips.
This development extends the role of T-Head’s chips across both major stages of AI computing: the initial training of sophisticated models and their subsequent deployment for user requests. Model Studio offers access to various models and tools, enabling the creation of AI applications.
Wu indicated that this internal coordination between Alibaba’s proprietary chips and its AI models is expected to yield improvements in computing costs and application development efficiency. He further suggested that the company’s comprehensive AI infrastructure, spanning both hardware and software, could offer a competitive edge within China’s domestic computing supply chain.
The integration of T-Head chips into the training of the Qwen model family connects two core areas of Alibaba’s AI investment: its foundational AI models and the hardware designed to support its cloud services. This strategic alignment aims to streamline the development and operational aspects of Alibaba's artificial intelligence initiatives.
分享這篇文章
相關文章
6 則
Anthropic's IPO bank closed position meant to stabilize stock, removing source of support
Apple develops Mac AI agent, plans third-party control safeguards - TechGig

Gemini 4 Neon Costs More Than Carbon Until December 1. Here’s Which Gemini Model You Should Actually Use

Hoymiles showcases versatile residential solar and storage solutions at RE+ 2026

How to deploy an app you built with AI to your own server

