ByteDance missed the 100-billion-yuan feast

43-year-old Zhang Yiming has just become the richest man in Asia.

The latest Bloomberg Billionaires Index shows that Zhang Yiming, founder of ByteDance, has a net worth of 105 billion US dollars, surpassing India's richest man Gautam Adani to top the list of Asia's richest people for the first time, ranking 18th globally.

But at the same time, in the global AI-driven wealth creation wave, ByteDance failed to become a winner.

Supported by hot capital, the valuations of star AI enterprises have risen rapidly, and the investors behind them have reaped huge profits. Since the beginning of this year, the two major rivals of ByteDance have obtained considerable unrealized book gains.

According to statistics from "Mirror Studio", as of early September, Alibaba's total unrealized gain from investment is about 1687 billion yuan, Tencent's is about 524 billion yuan, and even Meituan has about 199 billion yuan.

After model companies such as Zhipu AI and MiniMax received investments from Alibaba or Tencent, their share prices skyrocketed after listing, becoming the main source of book profits for large tech companies.

Both Alibaba and Tencent are keen to place bets on AI companies in advance.

In particular, Tencent has made frequent moves in the primary market in the past few years, taking stakes in a large number of AI startups. In many of the invested projects, Tencent is one of the largest shareholders.

With more and more star AI companies coming under its wing, Tencent has almost "bought" half of China's AI industry. Now these investments have begun to contribute tens of billions of levels of book returns.

In contrast, ByteDance hit the brakes on its investment five years ago, almost missing the entire AI track, which is a considerable misjudgment.

After 2018, ByteDance's investment once advanced by leaps and bounds, and the number of investments reached its peak in 2021. However, under the influence of various factors, in early 2022, ByteDance basically abandoned its strategic investment business on the grounds of "strengthening business focus and reducing low-synergy investments", leaving only sporadic financial investments.

During the same period, the global AI capital feast gradually kicked off, and ByteDance did not participate in it. On the whole, it insists on building the full AI stack on its own, with layouts ranging from models, Agents to Apps and programming software, and does not pay much attention to external investments.

ByteDance has become the only one among the three Internet giants that has not made large-scale investments in star AI projects.

However, ByteDance has rarely started to adjust its strategy recently.

Recent market sources say that well-known investor Kai Jiang has left Coatue to join ByteDance, in charge of the investment team based in Hong Kong, and will report to ByteDance CFO Gao Zun.

According to people familiar with the matter cited by Investment Press, Kai Jiang is in charge of ByteDance's financial investment department, which is not a newly established business, nor does it mean that the strategic investment department cancelled several years ago is restarted. From this point of view, Kai Jiang's rank and authority are not very high.

Even so, the introduction of an experienced investor who has a deep connection with ByteDance can still be regarded as a remedial measure after ByteDance realizes its mistake. It can even be said that ByteDance is going to learn from Tencent in terms of investment.

But time waits for no one, the first wave of AI investment focused on foundation models has drifted far away. Leading model companies at home and abroad have completed many rounds of financing, with high valuations and a certain degree of faction alignment.

Will ByteDance's quietly launched investment chase bring different big results from the past?

A

ByteDance is undoubtedly one of the main domestic AI players, with its products and businesses covering almost all mainstream tracks. But from the perspective of camp, ByteDance is also the most "lonely" AI player.

Up to now, ByteDance has very few allies in the AI industry, even barely any. In contrast, through continuous investment by Alibaba and Tencent, the "Alibaba Ecosystem / Tencent Ecosystem 2.0" in the AI era has begun to take shape.

Tencent's strategy is particularly eye-catching. In addition to leading companies such as the six AI dragon startups, DeepSeek and Manus, some less well-known companies, such as Yanyu Technology, also have the opportunity to get Tencent's investment.

The two giants waving checkbooks have a clear mind: by taking stakes in the "seed players" of the AI track, they can maximize their share of the dividends of the AI era. Even if their own AI business cannot be built up temporarily, they will not be completely stranded in the last era.

ByteDance certainly understands this truth, but it embarked on a completely different path very early.

As early as the end of 2021, Liang Rubo reorganized ByteDance into six major business segments, marking that this rising new giant made a key choice: between "building on its own" and "investing in others", it chose the former as its strategic direction.

Specifically for AI, ByteDance's "building on its own" strategy has achieved quite good results.

In the past five years, although the basic Doubao model has never reached the SOTA level, relying on Douyin's content ecosystem and traffic advantages, after ByteDance's full-stack AI is gradually rolled out, it still occupies a leading position in many aspects such as AI App and multimodality.

Winning consecutive victories in local battlefields is likely to greatly boost ByteDance's confidence in the "building on its own" strategy. In the past five years, a new batch of star enterprises have emerged in the AI field, with a development momentum far exceeding that of startups in the mobile Internet era; but ByteDance seems unmoved, still rarely making investments, and does not hold shares in star AI companies.

In one investment competition after another, ByteDance seems to choose to "watch the fire from the other side" — others try every means to invest, even at the expense of accepting harsh conditions, but ByteDance always stays out of it.

Behind this choice, there is of course the strategic consideration of adhering to "building on its own", but there is also a very realistic factor.

After ByteDance's strategic investment department was disbanded, no special person in the group's core management was responsible for this work. What has a more far-reaching impact is that ByteDance's investment function was delegated to business lines.

In the public response at that time, ByteDance stated that "employees of the strategic investment department will be dispersed to various business lines to strengthen the cooperation between the strategic research function and the business." In any large company, if the business department leads the investment, the amount is bound to be limited, and it must focus on "how to help the business", lacking the patience and tolerance for trial and error that investment requires.

Over the past five years, ByteDance's few investments have often focused on "how to assist Doubao".

It has successively invested in companies such as Moore Threads, Ziliang Robotics, Yingmu Technology and Yunmai Corelink, most of which follow this logic. Those projects whose prospects are temporarily unclear or directly compete with Doubao are difficult to enter ByteDance's investment scope.

But the development potential of such projects is often impressive. Today's DeepSeek, which is at the height of its power, started as no more than a research project of a quantitative fund. Manus, which returned to China not long ago, is also not aimed at serving the existing business of Internet companies.

Looking back now, the impact of cutting the strategic investment department in early 2022 is not only to shrink the scale and frequency of investment. It made the remaining investment force of ByteDance embark on the path of pragmatism, which runs counter to the long-termism naturally required by AI investment and the adventurous spirit of tolerating trial and error and failure.

B

Why did ByteDance, which is used to acting alone, suddenly introduce a head of financial investment?

Up to now, ByteDance has not given an official response.

But combined with the recent changes in the AI circle, a more reasonable explanation is: Alibaba, Tencent and other companies have made a lot of money by investing in AI companies, and ByteDance can't sit still anymore.

Since the beginning of this year, the AI capital market has become hotter day by day. In the primary market, the valuations of AI companies have multiplied, and early investors have made huge profits. In the secondary market, Zhipu AI and MiniMax have also created a lot of wealth miracles.

The AI wealth creation myth is still continuing. A week ago, Enflame Technology was listed on the Sci-Tech Innovation Board. After only a few trading days, its market value has exceeded 210 billion yuan. Tencent holds more than 20% of the shares, and this single investment alone has an unrealized gain of 40 billion yuan.

Faced with this situation, ByteDance cannot be unmoved.

However, the unrealized book gain brought by the rise in the market value of invested companies is only one of the values of giants investing in AI companies. In addition, as a major shareholder, it cannot sell at high points at will, so this part of the paper income is difficult to be realized and pocketed immediately.

The greater value is that giants can get technical and product "support" to ease the pressure on their own AI business.

Tencent has placed extensive bets in the field of foundation models, and very early stopped treating third-party models as rivals, but is willing to integrate them into its own products, standing side by side with the hy series models.

This move produced immediate results. Take DeepSeek as an example. At the beginning of last year, after Yuanbao was connected to DeepSeek R1, its daily active users surged by more than 20 times, once topping the free App Store list in China.

With Tencent taking a stake in DeepSeek, the relationship between the two companies has become closer. Up to now, DeepSeek is still one of the main models for Tencent's products such as WorkBuddy, and often provides exclusive discounts.

Helping cloud business lock in large orders by investing in model companies is another purpose of giants.

Alibaba has invested in five of the six AI dragon startups. MiniMax alone purchased about 75.9 million US dollars of computing power from Alibaba Cloud last year. From 2026 to 2028, this figure is expected to reach 300 million, 400 million and 500 million US dollars respectively. The remaining four companies also place a large number of training and inference tasks on Alibaba Cloud.

In addition, the brand bonus, ecological positioning and industry insight behind investing in star AI projects are also the benefits that giants have truly felt.

In contrast, ByteDance is at a disadvantage in this regard.

Doubao, which emphasizes "self-reliance", has its strong point in Apps. On the model side, although the Doubao series models have their own highlights, they have hardly reached the SOTA level of the whole industry.

When people talk about cutting-edge models at home and abroad, they either talk about new products from OpenAI and Anthropic, or pay attention to companies such as DeepSeek and Zhipu AI; the latter are more or less connected with Tencent and Alibaba, while ByteDance's name is rarely mentioned.

In the domestic AI cloud market, thanks to the Doubao App, Volcano Engine ranks among the top in terms of token call volume, but its revenue is less than that of Alibaba Cloud — the latter has a higher proportion of large customers, which can contribute more substantial revenue and profits.

Moreover, the opportunities of the AI era, in addition to taking the lead in realizing AGI, can also make a lot of money through investment. And the large company's self-development of full-stack AI is not opposed to actively investing in other AI companies.

Although Alibaba and Tencent have invested in a large number of projects, they still spare no effort in their own AI sectors. Foreign companies are the same: Google has the most complete full-stack AI self-research in the world, but at the same time it has also carried out large-scale strategic investments, including a commitment to invest a total of 430 billion US dollars in Anthropic.

ByteDance has been following the "building on its own" route for the past five years, and its performance is pretty good. But in the case of old rivals making money one after another, ByteDance has the willingness and necessity to change its strategy, rather than letting investment opportunities slip away in vain.

C

Stimulated by the two major rivals — especially Tencent — ByteDance's investment engine has slowly started to turn again.

ByteDance's last investment boom was also inseparable from Tencent's stimulation.

In 2018, with WeChat "blocking" Douyin, the "ByteDance-Tencent war" broke out, the two giants attacked each other, and Zhang Yiming even argued with Ma Huateng on WeChat Moments.

In the same year, ByteDance established a strategic investment department, trying to penetrate various segmented tracks through investment, and then build its own ecological moat. ByteDance, which is used to internal competition and acting as an App factory, began to move along the path that Tencent had taken.

In the following years, ByteDance made more and more investments. In 2021, ByteDance took stakes in about 60 companies, reaching the peak over the years. But this boom came to an abrupt end in 2022, and the number of investments shrank by 94% in the following three years.

Now, with the arrival of the AI era, ByteDance is going to learn from Tencent about investment again.

The difference is that the last time ByteDance learned from Tencent, it truly felt the pressure and threat from Tencent, and one of the purposes of making large-scale investments was to "use his own method to deal with him"; this time, Tencent did not directly put pressure on ByteDance. What ByteDance values is the rich returns and long-term value of Tencent "buying" half of the AI industry.

But compared with a few years ago, it is much more difficult for ByteDance to expand its camp again.

Up to now, star AI companies at home and abroad have completed multiple rounds of financing, and their market values have soared to tens of billions or even hundreds of billions of US dollars. In the primary market, such a high valuation can easily lead to late-entering capital failing to make money, or even facing long-term lock-up.

On the other hand, the biggest cake in the AI field — foundation models — has been almost completely divided up by the six AI dragon startups, DeepSeek and other players. They have absolute advantages in algorithms, computing power and capital, and almost all SOTA models come from them; other second- and third-tier companies can hardly shake the foundation model pattern.

This means that if ByteDance wants to capture promising startups, it cannot continue to focus on foundation models, but turn its attention to the Agent layer or application layer, such as various Agent frameworks, or AI Apps for specific fields. In terms of investment, it has missed the first wave of foundation model boom and needs to wait for the new tide to come.

Introducing Kai Jiang, starting from low-risk financial investment and conducting small-scale trials is a relatively safe path. It can be predicted that driven by Kai Jiang and his superior Gao Zun, ByteDance's name is expected to appear in more and more large deals.

At the same time, ByteDance is also exploring along another path — spin-off and financing.