How China’s need to win AI race drove the rehabilitation of ‘Crazy’ Jack Ma

“Crazy Jack” earned his nickname during the early years of Alibaba, the e-commerce giant he built into one of China’s most valuable companies. The unconventional corporate culture he created had staff doing handstands during breaks from punishing work schedules. He became China’s most famous businessman by far, jetting off to meet world leaders and indulging in cosplay on stage at Alibaba’s annual parties (his Michael Jackson act in 2017 remains a YouTube hit).

Ma fell out with Beijing’s mandarins not because of his audacious outfits but for daring to criticise them in public. He told a conference that China’s financial regulators lacked innovation and that state-owned banks had a “pawnshop mentality”. A crackdown followed, costing him his public visibility and perhaps at least half of his wealth.

In February, he got a second chance. Dressed in a sombre dark suit, Ma and other top tech bosses attended a meeting with Xi Jinping, China’s authoritarian leader. Although Ma did not speak, TV footage showed Xi shaking hands with him – a clear sign of at least partial rehabilitation.

Xi’s decision to bring Ma in from the cold was not just about Beijing’s desire to reinvigorate its private sector to drive economic growth, or Xi’s obsession with tech nology as a potentially decisive factor in China’s geopolitical rivalry with the US. It was, at its most essential, part of a quest to beat the US to artificial general intelligence (AGI), a so-far hypothetical level of AI in which machines can match, or surpass, humans in intellectual tasks. Just as the US-Soviet rivalry found expression in the space race, the superpower showdown between the US and China is reflected in efforts to forge ahead in AI.

Alibaba is recognised as a leader in AI, not just in China but globally.

“Jack is back, though he is still kind of on mute,” says Duncan Clark, author of Alibaba: The House That Jack Ma Built. “You won’t find him being outspoken these days, but he’s back in favour.

“Alibaba is a very serious AI player. Apple’s choice of Alibaba as its AI services partner for iPhone in China shows it’s a leading player in a highly competitive market. In this way, Alibaba is very much Xi’s kind of company.

“Xi isn’t so interested in the soft stuff. He wants deep science, and AI is definitely deep science.”

Alibaba, in which Ma remains one of the largest shareholders, is not shy about telegraphing its ambitions for AI. Eddie Wu, its CEO, describes AI as a “once-in-a-generation” opportunity, adding that AGI is the firm’s “primary long-term objective”, according to Alibaba, which recently said it would spend Rmb380bn ($53bn) over the next three years to advance cloud computing and AI infrastructure.

Alibaba, a mix of e-commerce, logistics, cloud computing and media businesses, aims to embed AI across its operations both to enhance customer experience and optimise efficiency. A global customer base of more than 1 billion people means that Alibaba already collects vast amounts of the data that is crucial for training AI on large language models (LLMs).

Jeffrey Towson, founder of TechMoat Consulting, says that this huge online user group allows Alibaba to test, train and market its latest AI tools, “so things are going to go faster than when you are selling your services to client after client”. In the titanic struggle for Asia’s cloud computing market, Alibaba, not Google or Amazon Web Services, is ahead.

Alibaba’s AI models, in contrast with its US rivals, are largely open source. In one sense, this chimes with Ma’s earliest ideas. Asked years ago how he chose the name Alibaba, he recalled asking a waiter in a San Francisco coffee shop what she knew about Ali Baba, and she replied: “Open ­sesame.” This fit, he said, as he wanted Alibaba to open the way for Chinese companies to sell stuff online.

Some of the AI models that Alibaba has developed under the Tongyi Qianwen name are open source. So influential is this technology that several of the top 10 open source LLMs in the world, as ranked by Hugging Face, are based on Alibaba “Qwen” models.

One advantage of an open source approach, as shown by the buzz about the Chinese AI ­sensation DeepSeek, is that it can help win acceptance for the technology as developers use its algorithms to ­create an army of AI agents. But darker questions lurk beneath the surface, including the extent to which China’s authoritarian worldview is baked into the algorithms driving AI agents to do a multitude of tasks in western democracies.

Di Dongsheng, an influential professor at Renmin University in China, argues that the foundation of LLMs lies in the values reflected in their algorithms. “Therefore, once we hold global leadership over ([AI)] models, we will, in effect, hold the most sway over global values and aesthetics,” he wrote in an influential paper. “The competition between China and the United States over the global market for large [AI] models is essentially a contest for dominance over global political ideology. For this reason, the issue of control over [AI] must be elevated to a matter of survival for the party and the nation.

It is this type of mindset that is giving Washington pause. The New York Times reported this month that the Trump administration and ­several lawmakers are looking into Apple’s deal to use Alibaba’s AI in its iPhones as they believe the deal would pose national security risks. Citing three people familiar with the issue, it ­­also said they worry that the tie-up between Apple and Alibaba would widen “the reach of Chinese chatbots with censorship limits”.

For Jack Ma, this may mean that as he emerges from the cold in Beijing, Alibaba’s global ambitions will meet a cooler reception abroad.

Photograph by CFOTO/Future Publishing via Getty Images