2026-09-22 · equities

Alibaba unveils new AI chip, calls it China's most powerful; shares gain

Source Investing.com UK Equities

(Singapore, 22.09.2026)Alibaba Group Holding is stepping up its challenge to global artificial intelligence leaders with a new homegrown AI chip and plans for a major expansion of data centres, as Chinese technology companies race to build more of the computing power needed for increasingly advanced AI models.

The Chinese technology giant on Tuesday unveiled the Zhenwu V900, an AI accelerator developed by its chip unit T-Head that Alibaba says delivers three times the performance of its predecessor. Chief Executive Officer Eddie Wu said the chips can be connected in clusters of as many as 500,000 units, allowing them to handle the enormous computing demands involved in training advanced AI models.

Alibaba described the V900 as China's most powerful AI chip, putting it more directly into a market dominated globally by Nvidia. The launch comes as US restrictions continue to limit Chinese access to some of the most advanced American-designed processors, encouraging China's technology industry to develop its own alternatives.

The scale of Alibaba's ambitions extends far beyond the chip itself. Wu said the company aims to build 20 gigawatts of global data centre capacity for Alibaba Cloud by 2032 as it prepares for what he described as "exponentially rising demand" for AI computing.

Alibaba has already committed more than US$53 billion to AI development over three years and raised about US$10.2 billion through a follow-on share offering in Hong Kong in August. The company has also indicated that it plans to list T-Head separately, according to Bloomberg News, potentially giving investors direct exposure to China's fast-growing AI chip industry.

Investors welcomed the latest announcement, with Alibaba shares jumping more than 5% in Hong Kong on Tuesday, while Tencent Holdings gained more than 7% as global technology stocks rallied.

From online shopping to AI infrastructure

Alibaba built its business around e-commerce platforms including Taobao and Tmall, but AI is increasingly becoming central to its long-term strategy. The company is investing across the entire technology chain, from chips and data centres to cloud computing and AI models.

Its Qwen family has emerged as one of the world's most widely used groups of AI models. Alibaba said in August that its open-weight models had recorded more than 3 billion downloads in six months, overtaking offerings from Meta Platforms, Alphabet and other Chinese developers to become the world's most downloaded AI model family.

Open-weight models allow developers to download and modify important parts of an AI system, making them easier and often cheaper to adapt for different industries. That strategy has helped Chinese models gain users at a time when many businesses are looking for lower-cost alternatives to proprietary American platforms.

Alibaba is now planning an even larger model with between 5 trillion and 10 trillion parameters, significantly bigger than many open-weight models available today and designed to perform longer and more complicated tasks.

All that development requires enormous amounts of computing power, which explains the company's push into both AI chips and data centres. T-Head has already delivered 560,000 processors from the Zhenwu family, with its chips deployed by more than 400 external customers, according to Alibaba. The company plans to update its AI chip range annually.

Like other Chinese chipmakers, Alibaba is focusing on connecting large numbers of processors into clusters to increase overall computing performance. Huawei Technologies and Moore Threads Technology are pursuing similar strategies.

The spending is putting pressure on Alibaba's finances. Quarterly profit plunged more than 75% in the latest quarter following a nearly US$10 billion investment push, but executives say the AI business is beginning to generate meaningful returns.

Annualised revenue from AI-related products is expected to reach US$10 billion in the current quarter, while management has said Alibaba should be able to recover its overall investment within three years. The company is also targeting annual cloud and AI revenue of US$100 billion within five years, about five times its current level.

China closes the AI gap

Alibaba's expansion is part of a broader shift in the global AI race, where Chinese developers are narrowing the performance gap with their US competitors despite having less access to cutting-edge chips.

Companies including Alibaba, DeepSeek and Moonshot AI have focused heavily on making models more efficient, allowing them to achieve strong performance while using less computing power. Chinese developers have also generally embraced open-weight technology, helping their models spread rapidly among developers and businesses.

Chinese AI models accounted for the largest share of generative model downloads on Hugging Face, a major platform for hosting AI models, according to Bloomberg. Some American companies, including Airbnb, DoorDash and Coinbase, have also adopted Chinese models hosted on local servers.

The growing popularity of Chinese AI comes despite continuing tensions between Washington and Beijing over technology. The US has restricted exports of its most powerful AI chips to China since 2022, citing national security concerns, while US officials have accused several Chinese developers, including Alibaba, of improperly using American AI models to help train their own systems through a technique known as distillation.

China has rejected the accusations as groundless and warned that it could respond if Washington acts against Chinese AI companies.

The dispute is gaining attention ahead of Thursday's meeting between US President Donald Trump and Chinese President Xi Jinping at the White House. AI is expected to be among the issues discussed, while US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng agreed this week to establish a dialogue aimed at developing a better understanding of AI-related goals and risks.

The competition is increasingly about more than which country can produce the smartest chatbot. AI requires huge amounts of electricity, data centre capacity, chips and other infrastructure, turning computing power itself into a strategic resource.

China has an advantage in its ability to rapidly add electricity generation and infrastructure, while regions with abundant renewable energy, including Inner Mongolia, have attracted hundreds of data centres used to train AI models. Beijing is also supporting some data centres with electricity subsidies when they rely entirely on Chinese chips.

Alibaba's planned 20-gigawatt data centre network illustrates how much larger the physical infrastructure behind AI is becoming. The company is betting that demand for computing power will continue rising sharply as AI spreads through businesses and industries, even as questions remain over how quickly the huge investments being made by both Chinese and American technology companies will generate sufficient returns.