Alibaba Zhenwu V900 took centre stage at Alibaba Group’s annual cloud conference this week as the Chinese technology company outlined a sweeping strategy spanning artificial intelligence models, custom silicon and global data centre expansion. Chief executive Eddie Wu said on Tuesday at the Apsara Conference in Hangzhou that the company plans to train a new AI model containing between five trillion and 10 trillion parameters, a step that would dwarf its current flagship and signal escalating competition in China’s AI sector.
Alibaba Zhenwu V900: Qwen 4 Training and the Push Toward ASI
Speaking to attendees, Wu said Alibaba’s Qwen team was continuing research into model architecture and data optimisation, with the stated goal of completing “more complex, longer-horizon tasks” and advancing toward artificial superintelligence, or ASI. In a separate statement, the company confirmed that its next-generation model, Qwen 4, was already in training. The upcoming Qwen 4.5 and Qwen 5 series are projected to scale up to between five trillion and 10 trillion parameters, compared with the 2.4 trillion parameters in the current Qwen 3.8 Max. Parameters are the variables a model learns during training and serve as a rough gauge of model size, meaning the planned releases would be roughly two to four times larger than today’s flagship.
Alibaba Zhenwu V900 Chip and Domestic Silicon Push
Alongside its model ambitions, Alibaba introduced the Zhenwu V900, a next-generation AI chip developed by its T-Head semiconductor unit. Wu described it as the most powerful AI chip in China, claiming it delivers three times the performance of its predecessor, the M890, which was released in May. A single cluster built on the new chip can support up to 500,000 cards for frontier model training and inference, according to the company. Mass production and commercial release are scheduled for the first quarter of 2027. The launch comes as Chinese firms race to build domestic alternatives to Nvidia’s processors amid US export restrictions, and Alibaba said it expects “significant growth” in annual AI chip shipments.
Alibaba Zhenwu V900: AI Supernodes and Infrastructure Bottlenecks
Wu noted that Alibaba’s proprietary M890 AI supernode already handles inference for models above two trillion parameters, a capability he said only “a handful of companies globally” possess. He added that customer demand for AI remained “exceptionally robust” and was accelerating Alibaba Cloud’s revenue growth, but acknowledged that global shortages across the AI data centre supply chain were limiting how quickly the company could expand. “The industry’s mid-to-long-term demand far outpaces our supply capabilities,” Wu said, adding that Alibaba Cloud would begin bringing its AI supernodes online at commercial scale this quarter.
Alibaba Zhenwu V900: Data Centre Expansion and Long-Term Targets
Looking further ahead, Wu set a target for Alibaba Cloud’s global data centre capacity to surpass 20 gigawatts by 2032, underscoring the scale of infrastructure investment required to support the company’s AI roadmap. The Qwen team has reported “meaningful progress” on recursive self-improvement, a process in which models identify their own limitations, design experiments and synthesise data to drive a cycle of self-evolution. Wu framed the current moment as the dawn of an era of “Machine Intelligence” comparable to the Industrial Revolution, predicting that machines would eventually produce more than 1,000 times the “thinking” of all humanity, up from less than 3 per cent today.
Alibaba Zhenwu V900: Bigger Picture for the AI Industry
Investors and analysts will be watching closely to see whether Alibaba can translate its ambitious targets into commercial results, particularly as US export curbs continue to shape the competitive landscape for AI hardware in China. Wu likened AI coding to the light bulb of the electrical age, describing it as an early application rather than the breakthrough product of the machine intelligence era. “The truly groundbreaking products of the machine intelligence era have not yet arrived,” he said. With the Zhenwu V900 on track for a 2027 release and the next Qwen series in development, Alibaba Zhenwu V900 and its surrounding ecosystem have become central to the company’s bid to compete at the highest tier of global AI development.
The Alibaba Zhenwu V900 announcement comes at a pivotal moment for China’s domestic semiconductor industry, as the company seeks to reduce its dependence on Nvidia GPUs that remain constrained by US export controls. By vertically integrating its own silicon with the next-generation Qwen model, Alibaba is positioning itself alongside rivals such as Baidu, Tencent and Huawei, each of which has pursued similar hardware-software optimization strategies. Analysts at Jefferies and Morgan Stanley have noted that custom AI accelerators can meaningfully lower inference costs when paired with proprietary models, potentially improving margins on cloud and enterprise AI services. The V900’s claimed performance specifications, if verified in independent benchmarks, would mark a significant step forward for Chinese-designed AI training chips, though details on manufacturing partner, process node and yield rates have not yet been disclosed.
Scaling a model to 5-10 trillion parameters would place Alibaba in direct competition with frontier systems from OpenAI, Anthropic and Google, while extending well beyond the company’s current Qwen3 flagship. Such a leap would require not only substantial compute infrastructure but also innovations in training efficiency, data curation and alignment techniques. Industry observers point out that parameter count alone does not determine capability, and that mixture-of-experts architectures could allow Alibaba to scale effective compute without proportional increases in training cost. Still, the capital expenditure implications are considerable, and the company’s recent commitment to multi-year AI infrastructure spending suggests confidence in long-term enterprise demand for large-model services through Alibaba Cloud.
The Apsara Conference unveiling also reflects a broader strategic shift toward monetizing AI through platform-level offerings rather than standalone consumer products. Wu’s framing of AI coding as a transformative general-purpose technology echoes similar rhetoric from Microsoft and Salesforce, both of which have emphasized AI-augmented software development as a near-term revenue driver. For Alibaba Zhenwu V900 to deliver on its promise, the company will need to demonstrate that domestic supply chains can sustain volume production, that developer adoption of Qwen APIs is accelerating, and that cloud customers view the integrated stack as a credible alternative to Nvidia-based solutions. Quarterly earnings reports from Alibaba Cloud in the coming fiscal year are likely to provide the first meaningful indicators of commercial traction.