Tencent has secured a dominant position in the AI industry through strategic investments.

In the field of AI applications, Tencent has also made intensive investments and started earlier. From 2019 to the present, it has taken stakes in companies such as Minglue Technology, Fanwei, and Fanshi Intelligence, covering mainstream tracks such as AI marketing, AI collaborative office, and AI enterprise platforms.

The latest exposed Tencent AI investment project is Yanyu Technology, the parent company of the well - known AI creation platform Liblib.

On June 18th, Yanyu Technology announced the completion of a $300 million Series B+ financing, co - led by Tencent, Shunwei Capital, and Granite Asia, with a valuation of over $2 billion.

As one AI star company after another joins Tencent's camp, Tencent has almost "bought" half of China's AI industry.

The strength and scale of these companies vary greatly. There is Zhipu, which has stirred up the large - model track and just exceeded HK$1 trillion in market value, and there are also many fledgling startups whose product technologies are yet to be tested by the market.

The key reason they can catch Tencent's eye is that they can help this giant quickly "make up for its deficiencies". In the fields of chips, computing power, models, and applications, if Tencent works alone behind closed doors, it may be difficult to achieve ideal results; by introducing external help, it can quickly make up for its shortcomings and achieve a much - needed overtaking on the curve.

The strategy of quickly forming competitiveness and building camp barriers with "money power" has been verified by Tencent in the mobile Internet era and achieved great success. Now, it is trying to replicate the same story in the AI era.

1

Tencent has been deploying in the AI field for a long time. In the past six months or so, as some of the invested companies have successively gone public, Tencent's long - term investment has begun to enter the financial return period.

In November last year, Minglue Technology, in which Tencent participated in multiple rounds of investment and held about 27% of the shares, went public on the Hong Kong Stock Exchange. Its market value exceeded HK$40 billion on the first day and has now dropped to about HK$35 billion. Based on this calculation, the value of Tencent's shares is about HK$9.5 billion.

Half a year later, another company in which Tencent holds shares and has a higher "AI content", Lightelligence, was listed on the Hong Kong Stock Exchange.

Lightelligence focuses on the field of optoelectronic hybrid computing power and is known as "the world's first AI silicon - photonics chip stock". Since April, its stock price has once reached HK$1,050 but then dropped to around HK$500, with a market value of about HK$48.5 billion.

According to Lightelligence's prospectus, Tencent holds 4.19% of the shares. Based on the current market value, the value of Tencent's shares is about HK$2 billion.

Although these two companies are supported by the AI concept, they are in relatively niche tracks; the two large - model star stocks are the "main courses" of Tencent's capital feast.

In early January this year, Zhipu and MiniMax were successively listed on the Hong Kong Stock Exchange, and their stock prices soared like rockets.

As of the close of the Hong Kong stock market on June 22nd, Zhipu has risen nearly 20 times in total, with a market value of up to HK$1.07 trillion. MiniMax has experienced a significant correction but still risen 274% in total, with a market value approaching HK$200 billion.

Tencent invested in Zhipu's Series B4 financing, spending 200 million yuan to obtain about 2% of the shares. Now, the book value of this investment exceeds HK$20 billion, a full 100 - fold increase.

On the other hand, around 2023, Tencent invested $50 million in MiniMax, equivalent to about 340 million yuan, holding 2.58% of the shares. Based on MiniMax's current market value, the value of Tencent's shares is HK$5 billion, an increase of about 14 times.

Since the end of the mobile Internet era, it has become increasingly difficult for domestic CVCs, and the return on investment has plummeted. However, Zhipu and MiniMax, with their amazing increases, have become quite rare classic cases of Tencent's investment.

A domestic AI chip company in which Tencent has invested is expected to go public soon.

On June 15th, Enflame Technology's IPO on the Science and Technology Innovation Board was approved by the Shanghai Stock Exchange's listing committee. According to Caixin, Tencent has invested in Enflame Technology five times, with a cumulative investment of 5.14 billion yuan. Compared with the shareholding ratios of one or two percentage points in Zhipu and MiniMax, Tencent and its concerted actors hold as much as 20% of Enflame Technology, ranking first.

Enflame Technology is known as one of the "Four Domestic GPU Dragons". The other three, Moore Threads, Muxi Co., Ltd., and Biren Technology, have all gone public, with market values ranging from HK$150 billion to HK$300 billion. If Enflame Technology reaches the same level, the value of Tencent's shares could reach HK$30 billion to HK$60 billion.

What is even more anticipated is, of course, Manus and DeepSeek.

Tencent took a stake in Manus early on. Meta, which entered the game in a hurry and left in a rush, raised Manus' valuation to $2 billion, which is also the rumored repurchase price.

According to the latest news, Manus is considering adjusting its corporate structure and setting up a Sino - foreign joint - venture enterprise registered in the country. Foreign media analysis suggests that this joint - venture structure is designed to allow Chinese investors to continue to hold the company's equity in a compliant manner and is also preparing for a future IPO in Hong Kong.

If it goes public in Hong Kong in the future, Manus' biggest selling points will be its team itself and its rapid growth rate. There are reports that in just a few months, Manus has increased its annual recurring revenue from $100 million to $400 - 500 million, comparable to companies like Zhipu.

As for DeepSeek, in early June, Tencent squeezed into the first - round financing, investing 10 billion yuan and becoming one of the largest external investors. Compared with its investments in companies like Zhipu, Tencent has increased the investment scale by two orders of magnitude.

Currently, DeepSeek's valuation has reached about 340 billion yuan, and Tencent's shareholding ratio is about 3%. Referring to the stock price trends of similar companies and considering the preference of the global market, DeepSeek has every chance to reach a market value of one trillion yuan; by then, the value of Tencent's shares is expected to exceed 30 billion yuan.

2

However, Tencent has never been satisfied with just being a financial investor. It always hopes to kill two birds with one stone, getting both financial returns and business and strategic support.

By taking stakes in other companies to quickly make up for the shortcomings of its own AI business is the biggest common feature of Tencent's AI investment.

Take AI chips as an example. As early as the end of 2021, Tencent launched three products - the AI inference chip "Zixiao", the video transcoding chip "Canghai", and the intelligent network card chip "Xuanling".

Among them, the Zixiao chip is widely used within Tencent and has a high cost - performance ratio in scenarios such as OCR text recognition, intelligent meetings, and image and voice noise reduction.

However, the Zixiao chip can only be used for AI inference, and it is difficult to compare with NVIDIA GPUs. It also lags behind domestic training - inference integrated chips such as Huawei Ascend and Baidu Kunlun. Tencent still needs to purchase a large number of GPUs externally to meet the computing power needs of its AI business.

On the other hand, Tencent is not certain about the high - cost project of self - developed chips.

In an open dialogue in early June, Tang Daosheng, the CEO of Tencent Cloud and Smart Industries Group, said when talking about self - developed chips that "designing chips by oneself does not solve the production capacity problem". Tencent chose to cooperate with more chip manufacturers and integrate with the ecosystem, making manufacturers willing to use Tencent as a benchmark for computing power display.

Enflame Technology is one of Tencent's main allies in the field of AI chips.

The two companies started to cooperate in 2019. According to Guancha.cn, Enflame Technology's first - generation training chip, Suisi 1.0, was tested and launched within Tencent Cloud, and Tencent's Zixiao chip is also the result of cooperation with Enflame Technology.

Meanwhile, Tencent is the main customer of Enflame Technology, contributing more than 80% of Enflame Technology's sales in 2025.

Enflame Technology is focusing on training - inference integrated chip modules, which have been taped out but have not yet been mass - produced and delivered. Compared with the Zixiao chip, which can only do inference, perhaps this is the heavy - weight product that Tencent values more and hopes to purchase in large quantities.

However, Tencent has not placed all its hopes on AI chips in one company.

In addition to investing in Enflame Technology, Tencent has also successively taken stakes in companies such as Yunbao Intelligence, Lightelligence, Jiyimei Semiconductor, and Wuwenxinqiong, which are distributed in the fields of DPU (Data Processor), photonic chips, digital - analog hybrid chips, and the IDC intermediate layer. Some are trying to open up new technology paths outside of GPUs and TPUs, some specialize in hardware around AI computing power, and some are conducting technological explorations for the future.

Basic large models are also a key area where Tencent is "calling for reinforcements".

Tencent's self - developed Hunyuan large model did not meet expectations before; the Hy3 Preview led by Yao Shunyu has improved significantly but is still in the stage of iterative upgrading. Fortunately, Tencent has successively taken stakes in the "Five Dragons", indirectly occupying a place in the field of SOTA models.

After becoming a shareholder of the "Five Dragons", Tencent has removed potential obstacles in the basic model for intelligent agent frameworks such as QClaw, Workbuddy, and Codebuddy - the sudden "supply cut" by Anthropic has revealed such risks.

Meanwhile, this move has bought more time and space for Tencent's own basic model R & D. Hy3 Preview does not forcefully pursue SOTA and ranking, and it is not fully promoted by Tencent, which has something to do with the group's various businesses being able to choose third - party models.

However, Tencent values DeepSeek more. It not only invested 10 billion yuan but also made DeepSeek one of the only two optional basic models for Yuanbao, standing side by side with Hunyuan. Compared with the "Five Dragons" that are busy pursuing PMF, DeepSeek, which focuses on AGI, is more in line with Tencent's long - term interests.

As for AI applications, Tencent has made relatively few investments.

Liblib is one of the invested companies. It is a startup focusing on AI image and video creation, claiming to have over 30 million cumulative users, generating over 5 million images per day on average, and having an annual recurring revenue close to $300 million. In addition, when Manus set off a general Agent craze, Tencent also took a stake and bet on its competitor Genspark.

Overall, Tencent's investment interest in AI applications is far less than that in AI chips and basic models.

After all, WeChat Agent is on the verge of launch. The Agent application ecosystem with WeChat as the entrance and mini - programs as the landing point is Tencent's development focus. Although AI services such as Liblib and Manus have their own advantages, their importance and strategic value cannot be compared with Agent - enabled WeChat.

Ma Huateng once described the challenges Tencent faces in AI as a "leaky boat". At that time, he did not elaborate on how to plug the leaks; but both insisting on self - development and "making up for deficiencies" in business through investment are obviously Tencent's long - standing coping strategies.

3

A few years ago, Tencent dismantled the "Tencent Ecosystem" through a series of equity clear - outs. Now, through a series of equity actions, Tencent has become a shareholder of a large number of AI companies, and the "New Tencent Ecosystem" in the AI era is on the verge of emerging.

AI star enterprises such as Zhipu and MiniMax, and domestic chip companies such as Enflame Technology and Yunbao Intelligence all have Tencent's presence behind them. Even when Manus was in trouble, Tencent was included in the rumored "problem - solver".

Meanwhile, AI entrepreneurs have also flocked to the "New Tencent Ecosystem", including many star players.

Wang Huiwen, a veteran of Meituan, founded Guangnianzhiwai, and Tencent invested $50 million. This is understandable: Meituan was originally a key player in the old Tencent Ecosystem.

But some entrepreneurs from Alibaba and ByteDance also choose to get money from Tencent.

After leaving Alibaba, Lin Junyang founded his own AI laboratory. In the budding state before the product direction was announced, he received $20 million from Tencent in the first - round financing. The other two "financial backers" were Gao Rong and Sequoia, and his former employer Alibaba was not among them.

Chen Mian, the founder of Liblib, previously served as the global commercialization director of Jianying under ByteDance. After leaving, he started a business. He received investment from Ming Shi Capital in the Series A financing, and money from Sequoia, CMC, and Ant Group in the Series B financing. Recently, he introduced Tencent as a new shareholder in the Series B+ financing.

Yanchuangwanwu specializes in AI intelligent agent programming. The two founders have respectively served as the algorithm director of TikTok and the technology product director of Baidu. It completed the Pre - A round of financing at the end of 2025, and Tencent invested tens of millions of dollars.

Founders with Alibaba or ByteDance backgrounds take money from Tencent at the start - up stage instead of seeking support from their former employers first. One of the reasons may be that Tencent offers more favorable investment terms than Alibaba and ByteDance.

Tencent has always been the most investment - savvy among the BAT giants and even the new BAT. This tradition dates back to the mobile Internet era, and now Tencent is trying to carry it over to the AI era.

But after the disintegration of the "Old Tencent Ecosystem", why does Tencent want to build a "New Tencent Ecosystem"?

Revitalizing the 400 - 500 billion yuan in cash on hand and finding a proper outlet for the money is the basic motivation for T