At its annual Apsara conference, Alibaba unveiled its evolving AI chip strategy, signalling a multi-year hardware development plan despite uncertainty over specific models and timelines, as US export restrictions push Chinese tech firms to accelerate domestic chip innovation.
Alibaba has stepped up its in-house AI chip push as Washington tightens export controls that have constrained Chinese access to Nvidia’s most advanced processors. At its annual Apsara conference in Hangzhou, the company said it is building a wider stack for AI, ranging from silicon and cloud infrastructure to models and agentic software, in a move that underlines how closely Chinese tech groups are linking chip design to their broader AI ambitions.
The main point of confusion in the reporting is which chip Alibaba actually introduced at the event. AP reported that the company unveiled the Zhenwu V900 and described it as three times faster than its predecessor, while other accounts said the newly shown part was the Zhenwu M890, with the V900 left as a future model on the roadmap. The common thread across the reports is that Alibaba is continuing to expand its custom accelerator line under T-Head, its chip unit, and is presenting those chips as suitable for both training and inference workloads.
What is clearer is the strategic direction. Alibaba has paired the chip announcements with broader plans for larger AI models and infrastructure. AP said the company is training a model that could reach 10 trillion parameters and aims to lift data-centre capacity to 20 gigawatts by 2032. That would place its AI effort among the more aggressive domestic responses to US restrictions, especially as Chinese firms seek alternatives to foreign accelerators.
The timeline for the V900, however, remains less certain. Some reports say mass production could begin in the first quarter of 2027, while others place the launch later in that year. Until Alibaba publishes a formal product roadmap or investor disclosure, the safer reading is that the company is signalling a multi-year hardware plan rather than committing publicly to a single launch date.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
