⚡Alibaba's Zhenwu M890 Hits 144GB, 3x Old Chip Speed
TL;DR
Alibaba's T-Head unit launched the Zhenwu M890 AI accelerator at a Hangzhou event May 20, with 144GB memory and 800 GB/s interchip bandwidth. The chip is 3x faster than the 810E and ships alongside Qwen 3.7-Max, a coding-and-agent flagship LLM.
Alibaba's T-Head unit launched the Zhenwu M890 AI accelerator at a Hangzhou event May 20, with 144GB memory and 800 GB/s interchip bandwidth. The chip is 3x faster than the 810E and ships alongside Qwen 3.7-Max, a coding-and-agent flagship LLM.
Key Points
Zhenwu M890 delivers 3x the performance of the 810E, with 144GB GPU memory and 800 GB/s interchip bandwidth
T-Head has already shipped 560,000 Zhenwu units to 400+ customers across 20 industries
Multi-year roadmap: successor V900 in Q3 2027, J900 in Q3 2028
Paired launch: Qwen 3.7-Max optimized for advanced coding and long-running agent tasks
Why It Matters
China just gave its biggest hyperscaler a credible Nvidia alternative for both training and inference, with real customer volume already shipped. Export controls become less binding as domestic silicon catches up on memory and bandwidth.
Quick Facts
Frequently Asked Questions
Why does this matter?
China just gave its biggest hyperscaler a credible Nvidia alternative for both training and inference, with real customer volume already shipped. Export controls become less binding as domestic silicon catches up on memory and bandwidth.
What happened?
Alibaba's T-Head unit launched the Zhenwu M890 AI accelerator at a Hangzhou event May 20, with 144GB memory and 800 GB/s interchip bandwidth. The chip is 3x faster than the 810E and ships alongside Qwen 3.7-Max, a coding-and-agent flagship LLM.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,337 builders reading daily.