Alibaba to boost AI spending as China tech giant sees AGI as new start
Group’s infrastructure spending will exceed the US$53 billion already promised, CEO Eddie Wu says at conference
Alibaba Group Holding plans to increase its capital expenditures on artificial intelligence infrastructure from the original 380 billion yuan (US$53 billion) over the next three years, the group’s CEO said on Wednesday.
Eddie Wu Yongming, also the chairman of Alibaba Cloud, said at the Apsara Conference in Hangzhou, in China’s eastern Zhejiang province, that Alibaba’s cloud unit aimed to become “the world’s leading full-stack AI service provider” from computing power to models. Wu did not, however, disclose the additional budget for AI infrastructure spending.
Wu also announced at a forum for international users of Alibaba Cloud the company’s biggest ever overseas investment in AI infrastructure with plans to launch the company’s first data centres in Brazil, France and the Netherlands, with additional data centers to be added in Malaysia, Dubai, Mexico, Japan and South Korea in the coming year. The investment amount for new data centres was also not disclosed.
“The overseas market is a very important market for us,” he said. “The growth rate of our overseas infrastructure expansion is far exceeding our domestic growth rate. We hope that in a few years, the scale of our international data centres will be five times what it was in 2022.”
According to Wu, Alibaba Cloud currently operates in 29 regions globally.
Alibaba’s shares gained 9.2 per cent to HK$174 on Wednesday. Alibaba owns the Post.
Wu’s pledge came as Alibaba is seeing surging demand for AI services and “an increasingly clear technology trajectory” from AI agents to AGI (artificial general intelligence) and ASI (artificial super intelligence) as he laid out the company’s “road map to ASI”.
AGI refers to a hypothetical type of AI that reaches or surpasses human cognitive capabilities, while ASI refers to an even more powerful future form of AI with intellectual scope far beyond human intelligence. US AI giants from OpenAI to Meta have all made achieving ASI an explicit goal.
“AGI will not be the end but a new beginning,” Wu said. The technology would save humans from 80 per cent of existing jobs, and the arrival of ASI would create “a group of super scientists and engineers” to solve problems from healthcare to climate change in “unthinkable ways”, he said.
He added that he was “optimistic” that the arrival of ASI would be good for humanity, revealing that the company has had discussions internally about how humans and ASI can coexist peacefully, at a time when some experts have warned that ASI could destroy humanity.
Wu’s pledge of additional investment also mirrored increased AI spending by US tech giants. US chip giant Nvidia said it would invest US$100 billion into OpenAI to help the US AI giant build and deploy at least 10 gigawatts of AI data centres for next-generation AI infrastructure.
In his speech to a room packed with developers and clients, Wu said the world was expected to spend US$4 trillion on AI computing infrastructure over the next five years. In the future, there would be “five to six” supercomputing platforms, he added, sending a strong signal that Alibaba – whose Qwen is the world’s most popular open-source model – would be one of the remaining few.
Wu said that AI large-language models would be the operating systems of the future, with AI cloud computing akin to the computer those operating systems run on.
By 2032, Wu expected the company’s global data centre energy consumption to be 10 times higher than 2022 levels.
“ASI will exponentially amplify the power of human intelligence, just like electricity did for humans’ physical capabilities,” he said. “In the past, 10 hours of work produced 10 hours of results. In the future, this may be multiplied by 10 or 100 times.”
The Apsara conference, an annual gathering of Alibaba Cloud users and developers, showcases hundreds of AI agents and real-world applications.
Zhou Jingren, the chief technology officer of Alibaba Cloud, promoted Alibaba’s latest offerings, from models to network infrastructure, including the HPN8.0, a high-performance computing platform for AI training and inference.
He said that Alibaba has open-sourced more than 300 AI models since the launch of the first generation of Qwen in 2023, with over 600 million downloads and more than 170,000 derivative models built upon Qwen.
In a separate panel event on Wednesday morning, executives from Industrial and Commercial Bank of China, NetEase Group and Li Auto discussed how they are integrating AI into their product offerings and harnessing the technology to generate economic value.
The conference runs until Friday. Major US chipmakers Nvidia, Intel and AMD are corporate sponsors.
AI industry faces US$800 billion revenue shortfall amid data centre expansion: Bain
By 2030, AI companies will need US$2 trillion in total annual revenue to fund the computing power needed to meet projected demand, Bain says
By 2030, AI companies will need US$2 trillion in combined annual revenue to fund the computing power needed to meet projected demand, Bain said in its annual Global Technology Report released on Tuesday.
The report is set to raise further questions about the AI industry’s valuations and business model.
However, the savings provided by AI and companies’ ability to generate additional revenue from AI are lagging behind that pace.
