Urgent.News

What's breaking now, across thousands of outlets.

AI

Alibaba unveils Zhenwu V900 AI accelerator, claims it's 'the most powerful AI chip in China'

Alibaba’s T-Head Zhenwu V900 claims three times the M890’s performance with 216GB of memory as Alibaba targets 10T-parameter Qwen

Alibaba unveils Zhenwu V900 AI accelerator, claims it's 'the most powerful AI chip in China'

Alibaba has introduced its new Zhenwu V900 AI accelerator and unveiled plans for the Qwen model at the Apsara Conference 2026, reports the AP. Alibaba CEO Eddie Wu claims the Zhenwu 900 to be the most powerful AI chip in China, boasting three times the performance of its predecessor, the M890. This chip is designed to handle the training of Qwen models with parameters ranging from 5 to 10 trillion, surpassing both the 2.4 trillion Qwen3.8-Max and Moonshot’s Kimi K3 models.

The Zhenwu V900 is equipped with 216 GB of GPU memory and offers 1,200 GB/s of inter-chip bandwidth. The chip can also scale out to a supernode cluster containing up to 500,000 cards. Alibaba plans to mass produce and commercialize the V900 in the first quarter of 2027, moving up from the previous timeline in May. As of May, T-Head, Alibaba’s chip arm, had shipped over 560,000 Zhenwu chips to more than 400 external customers, with an annual refresh expected.

The technical specifications of the chip do not include FLOPS, process node, foundry, power figures, or memory bandwidth claims. The M890, Alibaba’s previous chip, had 144GB of memory and 800GB/s of inter-chip bandwidth, supporting a supernode of 128 chips. Huawei’s upcoming Ascend 960PR is also slated for 2027, with 192GB of memory capacity, 2.4TB/s of memory bandwidth, and a 2.2TB/s scale-up interconnect.

The Zhenwu V900’s performance claims are based on potential maximum FLOPS, which may not match actual calculations, making real-world performance the ultimate judge. Alibaba aims to achieve a data center capacity of over 20 gigawatts by 2032, but global AI supply chain shortages are currently slowing down this scaling process. Despite these challenges, Alibaba expects a surge in its AI chip shipments domestically as Huawei postpones its global Ascend rollout.

The company also expects significant growth in shipments for its Qwen models, with Qwen 4 currently in training and Qwen 4.5 and 5 families planned for future releases.

Written by urgent.news from Tom's Hardware's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at tomshardware.com →

More in AI

More from Wednesday 23 September →