AI

Alibaba to Train AI Model With Up to 10 Trillion Parameters

Chinese tech giant also unveils its most powerful chip yet and sets ambitious data center expansion targets through 2032.

Omega Editorial· September 22, 2026· 2 min read

Alibaba Group is preparing to train a massive new artificial intelligence model containing between 5 trillion and 10 trillion parameters, CEO Eddie Wu announced Tuesday as the company outlined an aggressive expansion across AI infrastructure, chips, and cloud computing.

The announcement, first reported by Reuters, positions Alibaba to compete at the frontier of large language model development, where parameter count—the number of adjustable values that determine a model's behavior—has become a key benchmark of capability and scale.

New chip claims performance lead

Alongside the AI model plans, Alibaba unveiled the Zhenwu V900, which the company describes as China's most powerful AI chip. According to Alibaba, the V900 delivers three times the performance of its predecessor, though the company did not disclose specific technical specifications or comparative benchmarks.

The chip announcement comes as Chinese technology companies work to develop domestic alternatives amid ongoing restrictions on access to advanced semiconductors from U.S. manufacturers.

Massive data center buildout planned

Alibaba Cloud, the company's cloud computing division, set an ambitious target to exceed 20 gigawatts of global data center capacity by 2032. That timeline gives the company roughly six years to substantially expand its infrastructure footprint to support both its own AI development and customer workloads.

The capacity target reflects the enormous energy and infrastructure requirements of training and operating large AI models, which demand specialized computing resources at unprecedented scale.

Why it matters

Alibaba's announcement signals China's determination to compete in the global AI race despite semiconductor restrictions and geopolitical headwinds. A model in the 5-10 trillion parameter range would rival or exceed the largest models currently deployed by Western companies, though parameter count alone doesn't determine model quality or usefulness. The simultaneous push on chips, models, and infrastructure shows Alibaba building a vertically integrated AI stack—a strategy that could prove essential as competition intensifies and supply chains remain uncertain. For enterprises evaluating AI partnerships, the expansion of Chinese capabilities creates new options while raising questions about data sovereignty, regulatory compliance, and long-term platform risk.

The details were first reported by Reuters correspondents Liam Mo and Eduardo Baptista.

#alibaba#large language models#ai chips#data centers#china ai#cloud computing

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in AI

AI· 3 min read

AI Infrastructure Spend Could Hit $1 Trillion in 2026, Dimon Says

JPMorgan CEO sees hyperscaler investment more than tripling since last year, adding roughly 1% to GDP annually while raising inflation questions.

Via AI Watch · Sep 22, 2026
AI· 3 min read

Code for America Wins Google Grant for AI Tax Document Screening

The nonprofit will use the funding to automate document review in its GetYourRefund service, which has helped nearly 400,000 low-income Americans file taxes.

Via AI Watch · Sep 21, 2026
AI· 4 min read

Why AI Agent Evaluation Must Measure Task Completion, Not Just Tool Calls

Traditional LLM benchmarks break down when agents execute multi-step workflows across live environments—here's how evaluation frameworks are evolving.

Via AI Watch · Sep 21, 2026