Alibaba's Zhenwu V900 AI chip to enter mass production in Q1 2027

Alibaba's chip unit says the Zhenwu V900 triples the performance of its predecessor, with commercial release planned for the first quarter of 2027.

Alibaba's chip design unit T-Head has unveiled the Zhenwu V900, an AI training and inference processor the company says delivers three times the performance of the Zhenwu M890 it released in May.

Announced at Alibaba Cloud's Apsara Conference in Hangzhou, the Zhenwu V900 carries 216GB of memory and 1,200GB/s of chip-to-chip bandwidth. It natively supports FP8 and FP4, low-precision number formats that let chips run AI inference faster and at lower cost. Alibaba said the chip is scheduled for mass production and commercial release in the first quarter of 2027.

CEO Eddie Wu described the V900 as the most powerful AI chip in China. Alibaba did not publish benchmark details behind the threefold performance claim. It also did not disclose where the chip will be manufactured or on which process node.

“At Apsara Conference, we are introducing our next-generation AI chip, the Zhenwu V900. It is the most powerful AI chip in China today, delivering three times the performance of its predecessor, the Zhenwu M890. A single cluster built on V900 can support up to 500,000 cards to power frontier model training and inference. Backed by the proven maturity of T-head product lines and widespread adoption across our clients, we anticipate a significant growth in the annual AI chip shipment volumes,” Wu said during his keynote address at the summit.

Zhenwu V900 supernode and customer base

Alibaba also showed an upgraded supernode server that pairs the Zhenwu V900 with its own ICN Switch, Panmai SmartNIC and Zhenyue SSD controller chip. The company said the design can scale to a cluster of up to 500,000 cards.

T-Head's existing Zhenwu chips serve more than 650 customers in sectors including automotive, finance, energy and manufacturing. Wu expects significant growth in annual AI chip shipments, as Chinese firms work to build domestic alternatives to Nvidia's processors under US export restrictions.

Yitian CPUs for agentic workloads

Alibaba also outlined two CPUs designed for agentic AI tasks, both due in 2027. The Yitian 720 offers higher single-core performance, core density and energy efficiency than the Yitian 710. The Yitian 730 is the first CPU built on T-Head's own microarchitecture. Alibaba said it delivers up to 40% higher SPECint2017/GHz, a measure of integer performance per clock cycle, than the 710.

Alongside the chips, Alibaba said Qwen 4 is in training, with the Qwen 4.5 and Qwen 5 series projected at 5 to 10 trillion parameters. Its current flagship, Qwen3.8-Max, has 2.4 trillion parameters, Reuters reported.

The company also described a self-improvement experiment in which its Qwen3.8-Max model made more than 10,000 calls to electronic design automation (EDA) tools over 60 hours. The model produced chip bus modules with 42% less chip area and no loss in performance, according to Alibaba. The company did not say whether those modules are used in the V900.