The ambition and anxiety of Alibaba's AI investments
Alibaba has once again reaffirmed its investment in AI.
At the recently held 2025 Yunqi Conference, Wu Yongming, CEO of Alibaba Group, stated that large models are the next - generation operating systems, and AI cloud is the next - generation computer. Alibaba is actively promoting an artificial intelligence infrastructure construction plan with an investment of 380 billion yuan in three years. By 2032, the energy consumption scale of Alibaba Cloud's global data centers will increase tenfold.
Alibaba's substantial investment in AI is a microcosm of the increasingly fierce global AI competition.
Generative AI has triggered a new round of technological revolution. Microsoft is deeply bound with OpenAI, Google is fully promoting the Gemini model, and Amazon AWS is also continuously increasing its efforts in AI services. Since Tencent accelerated its AI strategy in the fourth quarter of last year, its capital expenditure in Q4 alone reached 36.578 billion yuan, a year - on - year increase of 386%...
Compared with other domestic technology companies, Alibaba is more aggressive in terms of investment and stronger in technology in this competition.
We try to figure out several key questions: Where exactly does the money invested by Alibaba in the AI industrial chain flow? How are Alibaba's "subsidiaries" doing? What changes have occurred in Alibaba's investment style in the AI era?
Where does the money from Alibaba's AI investment flow?
In the view of many investors, Alibaba's thinking in AI investment is "full - stack coverage": from computing power infrastructure to model research and development, and then to industrial applications, it tries to connect the entire chain as much as possible.
Investor Wang Ming explained that only with full - stack can AI be turned into a real infrastructure, enabling AI applications to be popularized in various industries with lower thresholds.
As he said, Alibaba has covered the entire AI industrial chain through the model of "self - research + investment + cooperation". The specific layout and representative cases are as follows:
Among the entire AI industrial chain (basic layer, technology layer, application layer), Alibaba has the heaviest investment in the basic layer.
At the latest quarterly earnings conference call, Wu Yongming said that in the past four quarters, Alibaba has cumulatively invested more than 100 billion yuan in AI infrastructure and AI product research and development.
Its subsidiary "T-Head Semiconductor" independently develops AI chips, and the "Damo Academy" tackles underlying algorithms and general AI technologies. Externally, it has invested in chip companies such as Cambricon and Yangtze Memory Technologies Co., Ltd. (YMTC).
For Alibaba, laying out the basic layer is not only to deal with the "chokepoint" risks but also determines whether it can take the initiative in technological iteration.
Secondly, it focuses on the most cutting - edge technology layer.
In addition to self - researching Tongyi Qianwen and the Feitian AI platform, since 2023, Alibaba has successively invested in large - model startups such as Zhipu AI, MiniMax, and Yuezhi Anmian. Recently, it has also bet on the unicorn in the field of AI video generation, Aish Technology.
Affected by the open - source of DeepSeek, large - model companies are gradually differentiating. Zhipu is betting on the Agent direction and continuously launching new SOTA models; Yuezhi Anmian is strengthening the long - text and multi - modal capabilities of its models; MiniMax's strategic focus is on multi - modal technology; Baichuan Intelligence has fully shifted to medical AI.
Through investments in multiple companies, Alibaba has covered different technological routes, so it can gain an advantage no matter which direction becomes the mainstream in the future.
It makes the most investments in the application layer.
In the field of embodied intelligence/robots, Alibaba is the most active. Robots can combine the "mind" of large models with the "body" in the physical world, making them the most ideal carriers for large - model capabilities to enter the physical world from the digital world. However, the technological route is not yet clear at present. Alibaba is seizing the track by investing in companies with different technological routes such as Unitree, Zhujidongli, and Zibianliang Robot.
Through investments in the application layer, Alibaba makes up for its shortcomings in specific AI technology fields and strengthens its scenario application capabilities.
With the full - industrial - chain layout, the list of Alibaba's external cooperation enterprises is constantly growing. A company that, due to its large business volume, considered building its own vertical - domain large model, decided after careful consideration to "let its own large model coexist with Tongyi Qianwen in the long term".
"Tongyi series has a wide variety of models, offering a larger selection range. Its open - source performance is good, and it can meet the needs more flexibly." In the view of the CTO of this company, Alibaba, which can make continuous high - level investments and stably provide high - quality products and services, is the best choice for the company's partner.
In addition to self - research and direct investment, the in - depth cooperation established by Alibaba with external enterprises constitutes a key part of its ecological expansion.
For example, Alibaba and Hainan Huatie jointly build a computing power hub and jointly establish a smart computing company; Alibaba Cloud and Lingyiwanwu establish an "Industrial Large - Model Joint Laboratory"; Alibaba Cloud's full - stack AI applications are used in automobile enterprises such as FAW, BYD, Geely, NIO, and XPeng, and penetrate into the mobile phone industry through cooperation with Honor.
These cooperations go beyond simple investment relationships, making Alibaba's AI industrial chain layout more three - dimensional.
How are Alibaba's AI "subsidiaries" doing?
Compared with Tencent, which has invested in industry giants such as JD.com, Pinduoduo, and NIO, Alibaba's investment in the AI field focuses more on technological ecological positioning rather than short - term returns. It is more like building a set of infrastructure for the future and integrating various capabilities into its own system.
Let's start with a special case. Among many invested enterprises, Unitree is one of the few that has achieved profitability.
In fact, before Alibaba invested in Unitree in 2025, Unitree had been profitable for five consecutive years since 2020. Wang Xingxing, the founder of Unitree, said in July that Unitree's annual revenue in 2025 had reached "about one billion yuan". Currently, Unitree has completed the share - reform and is expected to submit an IPO application before the end of 2025.
Although Unitree is "not short of money", it still chooses to raise funds because the R & D costs of core components such as high - torque motors and high - precision reducers are extremely high. By introducing industrial capitals such as Alibaba, Tencent, and Geely, it can not only obtain funds but also access industrial chain resources.
In addition to Unitree, there is a group of invested enterprises in the process of going public, including Smart Voice, YMTC, Zhipu AI, and Zebra Network.
According to public information, Smart Voice is a conversational artificial intelligence platform company that Alibaba invested in twice in 2015 and 2016. As a technology provider, Smart Voice applies voice interaction technology to the hardware products in the Alibaba ecosystem, such as the Tmall Genie smart speaker. Now it has provided voice interaction solutions for tens of millions of vehicles and has reached cooperation with 90% of mainstream home appliance brands.
At the beginning of this year, Smart Voice completed a 500 - million - yuan financing and is restarting its IPO, but it has not yet achieved profitability.
YMTC is an integrated memory manufacturing company that Alibaba invested in 2022. It started the IPO guidance in July. Benefiting from domestic substitution and capacity expansion, YMTC's market share has increased rapidly. According to Omdia data, its DRAM chip production is expected to increase by 68% compared with 2024 this year, exceeding 2.7 million pieces.
Zhipu AI received investment from Alibaba in 2023. It is currently continuously iterating its large model and actively promoting open - source. In April this year, it also started the IPO guidance and is expected to become the "first large - model stock" in China.
However, not all companies are going smoothly. Zebra Network, an intelligent cockpit enterprise jointly initiated by Alibaba and SAIC in 2015, is still deeply in the red. From 2022 to 2024, its cumulative loss exceeded 2.6 billion yuan. In the first quarter of 2025, its loss due to one - time impairment increased sevenfold year - on - year to 1.582 billion yuan.
In addition, since the proportion of the amount that Zebra purchased cloud services, software, and shared services from Alibaba from 2022 to the first quarter of 2025 all exceeded 50%, the outside world has questioned its business independence. On August 20 this year, Zebra submitted an application for listing in Hong Kong, but its prospects are still unclear.
The best example that reflects Alibaba's integration thinking is C-Sky Microsystems. In 2018, after Alibaba acquired this chip company, it integrated it with the self - developed chip business of the Damo Academy to establish T-Head Semiconductor. Since then, T-Head's self - developed chips have been deeply coordinated with Alibaba Cloud services to provide underlying computing power support for Alibaba's internal businesses.
Overall, the invested companies of Alibaba have different development directions. Some are accelerating their IPOs, some are still exploring in the face of losses, and some have been deeply integrated into the system. However, the investment logic behind them is the same, which is to concentrate superior resources, strengthen the competitiveness of the two core businesses of AI and cloud, and at the same time release the value of high - quality assets through capital operation.
That is to say, although some AI enterprises do not make money in the short term, they are currently in Alibaba's core cooperation and resource - inclined areas. If they can transform Alibaba's traffic, technology, and data into actual competitiveness, they are all "potential stocks".
"Taking large models as an example, some companies that receive investment from Alibaba pay part of the investment in the form of Alibaba Cloud services. The companies' data, traffic, and computing power needs will all feed back into Alibaba Cloud." In the view of technology investor Pan Guang, Alibaba is investing in the entire ecosystem rather than just the equity of these companies.
"The open - source and price - cut of Tongyi Qianwen are actually 'enclosing territory'. Through free and low - price strategies, it attracts developers and enterprises to use." Pan Guang said, "Once the ecosystem is formed, profits will naturally flow back through cloud services, solutions, and other means."
After all, Alibaba's ambition is not just to become an AI technology company, but to become the "water, electricity, and coal" of the entire AI ecosystem, and the foundation on which all AI applications depend.
Alibaba's investment approach has changed in the AI era
"Once Alibaba's 'blood' enters, it first instills Alibaba's values, aligns the granularity, sends in high - level executives, and then the company goes bankrupt."
"Never sell your company to Alibaba. If you have to, it's okay to take Tencent's money. Or Alibaba should learn from Tencent and just give money."
"A good company is going to be 'empowered' by Alibaba."
At the beginning of February this year, the news that "Alibaba plans to invest $1 billion in DeepSeek to acquire a 10% stake" spread in the market, triggering netizens to leave messages on social platforms. Subsequently, the news was denied. However, the netizens' attitudes reflect the impression left by Alibaba's previous investment style on the public.
Image source / pexels
For a long time, Alibaba's investment style has been compared with Tencent's. Tencent only seeks symbiosis in investment and does not seek ownership. Although it has been criticized for being "investment - bank - like", there are still many "outstanding companies" emerging. Alibaba, on the other hand, tends to conduct full - scale acquisitions or hold a large proportion of shares, incorporating the invested enterprises into its own territory. After being acquired, some enterprises have become "auxiliaries" of Alibaba's business, and their innovation ability and independence have been restricted.
Kaola.com is a typical example. In 2015, NetEase launched the cross - border e - commerce platform "NetEase Kaola". In the first half of 2019, Kaola led the industry with a 27.7% market share. In September of the same year, Alibaba acquired it for $2 billion and renamed it "Kaola.com".
After that, Kaola went downhill. In 2021, it was reported that the team had shrunk from more than 400 people to less than 20 people, and the supply chain, warehousing and logistics, and even the user group gradually shifted to Tmall Global. In March this year, the Kaola.com APP was taken off the shelves, and official operations came to a complete halt.
However, in the field of AI investment, Alibaba's investment strategy has been significantly adjusted.
In 2024, Alibaba invested in Yuezhi Anmian at an extremely fast pace. According to media reports, the key points of this deal are that part of Alibaba's investment is paid in the form of Alibaba Cloud services; Alibaba cannot prevent Yuezhi Anmian from accepting other investors; and Yuezhi Anmian has absolute decision - making power.
From this, we can see that Alibaba's investment strategy is changing in three aspects:
From "e - commerce ecosystem construction" to "AI infrastructure positioning". In the Internet era, Alibaba's investment focused on building a business ecosystem centered on e - commerce. Now, the purpose of investment has changed to consolidating and expanding the leading position of the "AI + cloud" infrastructure, that is, through investment, it binds with leading AI companies to ensure that its infrastructure does not lag behind the technological wave.
From "independent investment" to "joint investment". Previously, Alibaba tended to lead investments or directly integrate through acquisitions, emphasizing control. Now, in the face of tracks where single - round financing can easily reach hundreds of millions or even billions of dollars and the technological risks are extremely high, Alibaba has chosen to "huddle together" with other large enterprises and investment institutions for investment, sharing high risks and high costs.
From "strong - handed integration" to "respecting independence". Alibaba is no longer obsessed with "control" but gives the decision - making power to the invested companies. This is due to the high scarcity of AI talents and the uncertainty of technological innovation. Being too strong may instead stifle the creativity of enterprises.
Conclusion
On the AI track, domestic and foreign giants have launched a fierce confrontation.
Recently, Microsoft announced that it will invest $80 billion in building a self - developed AI chip cluster; Amazon AWS is building basic building blocks; Google will invest £5 billion in the UK in the next two years to build an AI economic system.
Image source / pexels
Domestically, Tencent plans to invest 100 billion yuan this year. In addition to purchasing high - end chips and building its own data centers, it focuses on the in - depth integration of "social + games". JD.com will continue to invest in the next three years to drive a trillion - level AI ecosystem; Meituan's R & D expenditure in Q2 this year was 6.3 billion yuan, a year - on - year increase of 17.2%, mainly due to investment in AI.