en.Wedoany.com Reported - At the 2026 Alibaba Cloud Summit held on May 20, Alibaba's semiconductor company T-Head comprehensively disclosed for the first time the commercialization achievements and product roadmap of its self-developed AI chip, the "Zhenwu" series. As of April this year, cumulative shipments of the Zhenwu AI chip have exceeded 560,000 units, serving over 20 industries and accumulating more than 400 customers.
T-Head Semiconductor Vice President Gao Hui elaborated on the chip's market performance at the summit. He stated that the Zhenwu chip is currently the most widely applied and fastest commercialized AI chip product in China. In the intelligent driving sector, over 130,000 Zhenwu chips have been deployed, compatible with more than 50 models, covering full-stack tasks such as perception, prediction, planning, and decision-making. Single-machine inference performance is improved by an average of over 50% compared to similar solutions, with partners including over 30 leading automakers such as Changan, BYD, and XPeng. In the financial industry, over 100,000 Zhenwu chips have been deployed, serving more than 150 banks, securities, and insurance institutions, supporting core scenarios like identity verification, document digitization, risk control management, and intelligent customer service. It also achieved the first domestic single-machine parallel inference of a trillion-parameter model. 
In terms of commercialization scale, T-Head's growth curve has become very steep. Alibaba Group CEO Wu Yongming previously revealed during a March earnings call that T-Head's annualized revenue scale has reached the 10 billion yuan level, and it is expected that the scale of high-quality AI chips it can produce will continue to expand from 2026 to 2027. Considering the cumulative delivery volume announced at that time was 470,000 units, the 560,000 units disclosed this time means its shipments increased by another 90,000 units in just two months.
External third-party data also corroborates its growth rate. According to a report released by market research firm IDC in April, in the 2025 Chinese cloud AI accelerator market, Huawei ranked first among domestic manufacturers with shipments of 812,000 units, while T-Head jumped to second place with 265,000 units shipped, demonstrating the strong capabilities of internet giants in the underlying hardware sector. The latest shipment figure of 560,000 units indicates that its growth rate may further accelerate, reflecting the market's urgent demand for domestic AI computing power.
Beyond commercial achievements, T-Head also publicly revealed its ambitious product roadmap for the first time. Gao Hui announced that the Zhenwu chip will enter an "annual iteration" cadence. Following the launch of the new generation training-inference integrated AI chip Zhenwu M890 (equipped with 144GB of memory, 800GB/s inter-chip interconnect bandwidth, and 3x the performance of the previous generation Zhenwu 810E), T-Head plans to launch the Zhenwu V900 chip in Q3 2027, which will again deliver a 3x performance improvement, featuring 216GB of memory and 1200GB/s inter-chip interconnect bandwidth. Furthermore, the Zhenwu J900 chip is scheduled for release in Q3 2028, achieving a leapfrog innovation in its self-developed parallel computing architecture. Additionally, the summit saw the release of the Panjiu AL128 supernode server based on the Zhenwu M890, equipped with the self-developed ICN Switch 1.0 interconnect chip, enabling efficient collaboration of 128 AI chips with P2P communication latency below 150ns, primarily targeting massive concurrent inference and large model training demands in Agent scenarios.
Gao Hui further emphasized that T-Head is currently one of the few domestic companies with the capability to cover full-stack data center chips. Its product line spans the Zhenwu series AI chips, Yitian series Arm server CPUs, Panmai series smart NICs, Zhenyue series storage controller chips, and ICN Switch interconnect chips, forming a comprehensive self-developed capability from computing power and storage to networking.
Regarding the IPO topic, Alibaba Group CEO Wu Yongming previously stated that T-Head does not rule out the possibility of an independent listing in the future, but there is currently no clear timetable. As its product iteration gets on track and commercialization scale grows rapidly, T-Head's capital path as an independent semiconductor company has become increasingly clear. From IDC data to the newly announced 560,000 unit shipment volume, within a year, T-Head has not only established its market position as the second-largest cloud AI accelerator supplier in China but also presented the market with a complete AI computing power blueprint spanning chips, complete machines, and full-stack data center architecture.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
