Alibaba's T-Head has deployed more than 100,000 of its self-developed Zhenwu AI chips across 150-plus financial institutions in China, including banks, brokerages, insurers and fund companies. Cumulative shipments have reached 560,000 units, with the chips handling wealth management, credit risk control, investment research and compliance monitoring. The company plans to launch the more powerful Zhenwu V900 and J900 chips over the next two years.
Alibaba's T-Head has deployed more than 100,000 self-developed Zhenwu AI chips across 150-plus financial institutions, marking one of the largest domestic chip rollouts in China's banking sector.
"This deployment scale shows AI chips are moving from pilot to production in financial services," Zhang Chi, Vice President of the Public Cloud Business Unit at Alibaba Cloud Intelligence Group, said at the 32nd China International Financial Exhibition.
Cumulative shipments of the Zhenwu series have reached 560,000 units, T-Head first disclosed at the 2026 Alibaba Cloud Summit last month. The chips are deployed in wealth management, credit risk control, investment research and compliance monitoring across more than 150 banks, brokerages, insurers and fund companies.
The ramp strengthens Alibaba's position in China's AI chip market, where domestic alternatives to Nvidia's H100 and B200 are in high demand following US export restrictions on advanced semiconductors. T-Head plans to launch the more powerful Zhenwu V900 and J900 chips over the next two years, a roadmap that could extend its reach beyond finance into broader cloud and enterprise AI workloads.
The deployment gives Alibaba Cloud a competitive edge against Huawei's Ascend chip series, the other major domestic AI chip contender in China. While Nvidia's H100 delivers 990 TFLOPS of FP16 performance per published specifications, Alibaba has not disclosed comparable benchmark data for the Zhenwu series against either Nvidia or Huawei products.
Alibaba's in-house chip strategy mirrors moves by other hyperscalers. Amazon's Trainium and Google's TPU have similarly reduced reliance on external GPU suppliers for internal workloads. For Alibaba, the Zhenwu chips also support its Qwen large language model family, creating a vertically integrated AI stack from silicon to application.
The financial sector represents a lucrative beachhead. China's banking industry spent an estimated 300 billion yuan ($41.4 billion) on technology in 2025, according to IDC, with AI infrastructure accounting for a growing share. T-Head's existing relationships with 150-plus financial institutions provide a distribution channel that rivals such as Cambricon Technologies and Biren Technology have yet to match.
Alibaba Cloud does not disclose pricing for the Zhenwu chips, which are offered as part of its cloud services rather than as standalone products. The company said cumulative shipments of 560,000 units span both internal Alibaba workloads and external customer deployments.
For investors, the chip ramp could improve Alibaba Cloud's margin profile over time. The cloud unit reported revenue of 31.7 billion yuan ($4.4 billion) in the December quarter, up 13% year over year, with AI-related revenue growing at triple-digit rates. Replacing Nvidia GPUs with in-house silicon for inference workloads could reduce cost per query by 40% to 60%, according to estimates from Bernstein analysts. Alibaba's US-listed shares trade at roughly 12 times forward earnings, a discount to Amazon's 35 times, partly reflecting the market's uncertainty about the pace of Alibaba's AI monetization.
This article is for informational purposes only and does not constitute investment advice.
