EL7.AI
StocksMediumUpdated×8•Originally published 22 September 2026•Updated 23 September 2026•
2 min read

Alibaba Unveils Zhenwu V900, Targets More Than 20 GW of Data-Center Capacity by 2032

Key Facts

1Alibaba says the Zhenwu V900 delivers 3 times the performance of the Zhenwu M890 and set Q1 2027 for mass production and commercial launch.
2Alibaba Cloud targets more than 20 GW of globally operated data-center capacity by 2032.
3BABA closed at $116.31 and BABAF closed at $14.45 on September 22, 2026.

Alibaba's T-Head unit unveiled the Zhenwu V900 AI processor during the Apsara Conference in Hangzhou, held from September 22 to 24, 2026. The company says the chip delivers 3 times the performance of the Zhenwu M890, while Alibaba Cloud targets more than 20 GW of globally operated data-center capacity by 2032.

According to the company announcement, the Zhenwu V900 has 216 GB of memory and 1,200 GB/s of inter-chip bandwidth, with support for FP8 and FP4. Alibaba scheduled mass production and commercial release for Q1 2027 but did not disclose detailed benchmark methodology or the workloads used for the 3-times performance comparison.

Alibaba positions the processor within a full-stack strategy spanning foundation models, proprietary chips and cloud infrastructure. In March, the company said T-Head processors had entered scaled production and support training, fine-tuning and inference.

More memory and faster interconnects can keep more model data close to processors and ease transfer bottlenecks, potentially raising throughput and lowering inference costs if the claimed gains are achieved. The 20 GW target measures electrical data-center capacity, not a direct benchmark of computing performance.

BABA closed at $116.31 on September 22, 2026 after trading between $116.06 and $120.40, while BABAF closed at $14.45 within a $13.44-$14.80 range in the same session. Building that capacity raises capital intensity before revenue is realized, so the plan's investor significance depends on delivering the Zhenwu V900 and converting data-center investment into cloud-revenue growth.

Latest Updates · 4

  1. Notable·

    Update: Alibaba further revealed that its upgraded supernode architecture is capable of supporting massive clusters of up to 500,000 cards. Alongside this infrastructure scaling, the company confirmed that its next-generation AI model, Qwen 4, is currently in the training phase, signaling a direct application for the newly announced hardware capabilities.

  2. Notable·

    Update: Alibaba has officially branded the new processor as the 'Zhenwu V900', with the company describing it as the most powerful AI chip in China. This branding extends to the broader Zhenwu series, which also includes the predecessor M890 model.

  3. Notable·

    Update: Alibaba plans to train a new AI model to bolster its competitive position, coinciding with scheduled discussions between President Donald Trump and President Xi Jinping regarding the global AI race this week. This move links the company's technical roadmap to broader geopolitical dynamics that could influence semiconductor supply chains and international technology standards.

  4. Notable·

    Update: Alibaba has officially described its new AI chip as the most powerful in China, aiming to solidify its position as a domestic tech leader. This claim underscores the company's ambition to outperform local rivals and close the gap with global technology standards.

Open on chart