Alibaba is expanding its activity across all layers of the artificial intelligence technology stack; from language models and software to chips and cloud infrastructure. The company, which began as one of China’s largest e-commerce players, has now made heavy investments in AI and intends to supply a significant portion of its required processing capacity with proprietary technologies.
According to ITNA, Alibaba introduced its new AI chip named Zhenwu V900 and described it as its most powerful AI accelerator. Published information says this chip is designed to run heavy AI workloads, including training advanced models.
However, Alibaba’s current official information about the Zhenwu M890 chip has been released, which the company described as its most powerful AI processor at the time of introduction. Alibaba had stated that the M890 delivers up to three times the performance of its previous generation, the Zhenwu 810E, and features 144 GB of on-chip memory and 800 GB/s inter-chip bandwidth. The chip also supports computations from FP32 down to FP4.
T-Head, Alibaba’s chip design arm, is responsible for developing the Zhenwu family. The company also introduced the ICN Switch 1.0 network chip, which offers up to 25.6 Tb/s bandwidth and enables connecting 64 accelerators with high-speed communication. This set of hardware, together with the T-Head SAIL software, is designed to create an integrated processing infrastructure.
The importance of this program becomes clearer when we know that Alibaba intends to sharply increase the capacity of its data centers and cloud services to meet the growing demand for AI. The company’s CEO, Eddie Wu, previously announced Alibaba’s plan to invest more than $53 billion in developing AI infrastructure and cloud computing over several years.
Alibaba is also increasing the commercial use of its proprietary chips. According to the company, by April 2026 more than 560,000 Zhenwu chips had been produced and shipped, and over 400 external customers across more than 20 industries were using these chips. In later Alibaba reports, the number of customers using the M890 also exceeded 650 companies.
This program is not limited to hardware. Alibaba has simultaneously developed the Qwen model family and its cloud infrastructures. For example, the Qwen3.7-Max model, designed with a focus on agent-based tasks, was able in Alibaba’s internal testing to run continuously for 35 hours and perform more than a thousand tool calls.
Alibaba’s extensive investment in this sector is happening while demand for AI compute power is rapidly increasing worldwide. Large tech companies need massive data centers and specialized accelerators to train and run advanced models, and consequently, designing a proprietary chip can give them more control over cost, capacity, and the supply chain.
In its recent financial report, Alibaba also announced that revenue from cloud and AI-powered computing services for the quarter ended June 2026 grew 45 percent to $7.1 billion. The company also stated that revenue from AI-related products has seen triple-digit growth for the twelfth consecutive quarter.
According to Alibaba’s official information, the company intends to use a combination of proprietary chips, Qwen models, and cloud infrastructure to create a full AI stack; an approach that could bring this company’s competition with major players in the chip and cloud computing markets to a new stage.