Today, we're diving into the fascinating AI boom that's been sweeping across China since early 2025, examining what this means for the global AI landscape and markets.
Just hours after the launch on March 6th of Manus, a Chinese AI agent, its registration site crashed from the sheer volume of visitors. Butterfly Effect, the company behind Manus, boldly claims its technology outperforms OpenAI's models, though they're now forced to grant previews by invitation only as they struggle to handle the traffic. There are even reports of registration codes being sold on the black market.
Manus is merely the latest manifestation of the AI mania that has swept over China since January, when DeepSeek – which I covered back in Episode #860 – shook the global AI community with a model that delivered comparable performance to Western counterparts at a fraction of the training cost. The effect on Chinese markets has been nothing short of staggering. Chinese stocks are experiencing their best start to a year on record, outpacing American ones by a considerable margin. The Hang Seng Tech Index, which tracks the biggest Chinese tech companies listed in Hong Kong, has surged by more than 40% since mid-January alone.
The excitement stems from a belief that more affordable AI will help innovators develop novel applications for the technology. Cloud computing providers are ramping up investment in data centers, triggering a cascade of capital spending throughout the supply chain. But the big question remains: will this boom have staying power?
In recent weeks, we've seen hundreds of large Chinese enterprises – spanning sectors from automotive manufacturing and state-owned energy companies to banks and food-and-beverage distributors – announce plans to integrate DeepSeek's technology. Even tech giants like Tencent are embedding DeepSeek models into their products, despite having developed models of their own. This widespread adoption extends to the public sector too, with city governments integrating DeepSeek's models into mobile applications for basic citizen services, while government departments, hospitals, and universities across China are exploring how to employ it for what they call "party building" activities.
The hype has reached such a fever pitch that local equity analysts joke that they must find a DeepSeek angle if they want their research reports to gain any traction. Some investors have gone so far as to speculate that DeepSeek could single-handedly revive the property market in Hangzhou, where the company is headquartered.
Chinese venture capitalists are equally enthusiastic. One Beijing-based VC noted that integrating DeepSeek's technology into her portfolio of robotics companies has led to significant cost reductions and performance improvements. Amid this excitement, countless AI startups have emerged across China. Some venture investors are pouring money into these startups even while acknowledging they're likely witnessing a bubble forming. As one Hangzhou-based investor put it: "It's overwhelming but we have no other choice. The economy is not good and there's not many opportunities elsewhere. So we have to go into AI as fast as possible." The strategy, according to this investor, is to invest in an "A" round – the earliest financing series – and exit during an "A+" round, which might come just a few months later.
The Chinese government is also leaning into this trend. On March 6th, China's central government announced plans to establish a venture capital fund with a staggering trillion yuan (approximately $140 billion) earmarked for tech-focused investments.
China's tech giants – including Alibaba, Baidu, Huawei, and Tencent – are embracing the hype and hoping to capitalize on the boom, particularly through their cloud computing divisions. Last month, Alibaba made the ambitious proclamation that its main objective was to achieve human-like artificial general intelligence. And, on March 6th, it released a new reasoning model that it claims matches DeepSeek's capabilities.
Alibaba has committed around $53 billion over the next three years to build data centers to meet the surging demand for AI cloud services – more than the company spent over the past decade. As the market leader in China's cloud sector with a 36% share, Alibaba appears to be betting that growth in this area will offset the sluggishness in its core e-commerce business. Baidu has already experienced a substantial increase in its cloud revenue, helping to counterbalance declines in other divisions.
The hardware side of this boom is equally impressive. Demand for servers optimized for AI has skyrocketed since early February, roughly coinciding with DeepSeek's rise to prominence. Suppliers have begun offering "all-in-one" servers that come pre-equipped with AI software. Many of these are sold directly to companies, including state-owned enterprises, that prefer on-premises servers for enhanced security. Sangfor Technologies, founded by former Huawei employees, has been one of the biggest beneficiaries of this trend, with its share price surging by approximately 140% so far this year.
China's AI boom is driving capital investment throughout the country's hardware supply chain. According to analysts at Jefferies, server manufacturers may spend more than 1.4 trillion yuan over the next two years as they expand production capacity.
However, some analysts are beginning to urge caution. Kai Wang of Morningstar argues that DeepSeek won't fundamentally change most of the companies that have benefited from the recent stock market rally in China. Another recent rally faded when anticipated strong government support for the economy failed to materialize. The same could happen this year, Wang suggests, if companies struggle to monetize AI effectively.
Access to advanced semiconductors could be another challenge. For now, the supply appears sufficient. Companies can still purchase H20 chips from Nvidia, though these are less powerful than Nvidia's top-tier chips that America has barred China from buying. Local chip designers like Cambricon, Enflame, and Huawei are working to close the gap and have already begun supplying some Chinese AI firms.
Yet semiconductor limitations could still cause China's AI frenzy to lose momentum. Some analysts worry that as new applications emerge, driving demand for ever more computing power, constraints on chip supply will become increasingly problematic. China's leading foundry, the state-owned SMIC, faces serious capacity constraints and can't produce the most advanced semiconductors. Even Huawei's best locally designed chips still lag significantly behind Nvidia's in performance.
The geopolitical dimension adds another layer of uncertainty. The Trump administration is reportedly considering harsher restrictions on China, including limiting access to H20 chips. China's market rally is predicated on the belief that the cost of training and running AI models will continue to decline. By restricting access to chips, America could potentially drive those costs back up, bringing China's AI euphoria to an abrupt end.
As with many technological trends, we're seeing both genuine innovation and speculative excess in China's AI sector. The long-term implications for global AI development, market dynamics, and geopolitical relationships remain to be seen. These are indeed fast-moving times and we'll continue to monitor these developments, although this is one of the first episodes I’ve done exclusively on AI’s geopolitical impacts. If that’s something you’re into hearing all the time, I highly recommend the news podcast Last Week in AI.
The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.