Is the next step for Doubao to undergo a spin-off?

Recently, a short internal message was announced within ByteDance's Seed team. The latest price of the team's incentive "Doubao shares" has been adjusted to $14.85.

Two months ago, ByteDance initiated the first repurchase of Doubao shares at a price of $13.08. Even earlier, the grant price of Doubao shares was $10.

This means that the value of Doubao shares has been re - evaluated twice in just over two months. In April, the price of Doubao shares increased by 30.8% compared to the grant price; in June, it rose by 13.5% compared to the repurchase price in April. During the same period, the price of ByteDance's options increased from $229.5 to $235.5, a rise of 2.63%.

In the past, the price of ByteDance employees' options mainly followed the changes in the group's valuation, supported by multiple businesses such as Douyin, TikTok, e - commerce, advertising, and enterprise services. The emergence of Doubao shares means that ByteDance's large - model business now has its own internal options system.

For employees of the Seed and Doubao businesses, personal earnings are now tied to the growth of the AI business. For the external market, Doubao is being valued separately from ByteDance's vast business portfolio.

Almost at the same time, Doubao's paid plan emerged.

At the beginning of May, three subscription plans appeared on Doubao's App Store page: the standard version at 68 yuan per month, the enhanced version at 200 yuan per month, and the professional version at 500 yuan per month. Doubao later responded that free services would continue to be provided, and value - added services were still in the testing phase.

These two actions together point to two changes in ByteDance's AI business.

The price increase of Doubao shares represents a change at the organizational and valuation levels. The testing of the professional version represents a change at the revenue and commercialization levels.

Against this backdrop, discussions about "whether ByteDance will spin off Doubao" are increasing. Especially since competitors have provided examples.

In May, Kuaishou announced that its board of directors was evaluating the restructuring of its Keling AI - related assets and business and might introduce external financing. Subsequently, Kuaishou's first - quarter financial report revealed that Keling AI's revenue in the first quarter exceeded 650 million yuan, a year - on - year increase of over 300%; the annualized recurring revenue in March was close to $500 million.

Keling provides a capital reference for the AI business of platform companies: the AI business can remain within the group for continued investment, or it can be spun off for separate financing and valuation.

Doubao and the Seed team are in a similar situation.

It has an independent App entrance, the Seed team and the Doubao large model, and scenarios provided by ByteDance's ecosystem such as Volcengine, Douyin, CapCut, e - commerce, and search. It has also started to establish its own pricing system through Doubao shares and paid testing.

However, behind the concept of spinning off Doubao is a large AI business team that connects models, computing power, cloud services, content ecosystems, and commercialization scenarios. If ByteDance spins off Doubao, it will reshape the boundaries of the company's entire AI landscape.

1

When discussing whether Doubao has the foundation for a spin - off, the example provided by Keling is clear. Large companies are willing to evaluate the restructuring of their AI business usually when two conditions are met simultaneously: the AI business requires continuous large - scale investment, and commercialization has achieved initial results.

Keling just meets these two conditions. The latest public data shows that in the first quarter of this year, Keling AI's revenue exceeded 650 million yuan, a year - on - year increase of over 300%; the ARR disclosed in March was close to $500 million.

Information disclosed in Kuaishou's financial report and by its management shows that Keling's revenue mainly comes from two sources: the application programming interface (API) calls of enterprise customers (i.e., the interface for software to call each other's capabilities) and paid membership subscriptions from end - users. The number of Keling's global users, enterprise customers, and paid users on the App has all increased rapidly in the past year.

The core reason for Kuaishou to evaluate the restructuring of Keling AI - related assets is that Keling has presented revenue data. The capital market can see the potential for it to realize its value, which is the first threshold for the spin - off of an AI business.

On the other hand, Kuaishou's capital expenditure is expected to increase significantly in 2026. The new investment will mainly be used for the procurement of servers, the expansion of data centers, and other AI computing power infrastructure.

Obviously, the pressure from the cost side is another important reason for Kuaishou to spin off Keling.

This provides a more realistic reference for Doubao. Doubao is approaching the same juncture, but it has a larger user base and greater cost pressure.

QuestMobile data shows that in March 2026, Doubao's monthly active users reached 345 million, ranking first among AI - native apps. In the first quarter, the average number of monthly uses per user of Doubao was 54.8, higher than that of DeepSeek and Qianwen.

At the model level, data disclosed by Volcengine shows that the daily average Token calls of the Doubao large model exceeded 120 trillion in March 2026, doubling in the past three months.

Behind the large user base, a more acute contradiction lies in the revenue level.

LatePost reported that as of the first half of this year, more than 200 million people used the Doubao app every day, but the daily revenue of the Doubao app was less than one million yuan, mainly from e - commerce commissions.

By May this year, based on the public API price of Volcengine, the gross profit margin of the Doubao large model, and user usage habits, it was estimated that the daily consumption of the Doubao app reached tens of millions of yuan.

Doubao's move towards paid testing occurred at this point.

At the beginning of May, three paid plans appeared on Doubao's App Store page: the standard version at 68 yuan per month, the enhanced version at 200 yuan per month, and the professional version at 500 yuan per month.

Public reports mentioned that the paid services are mainly targeted at professional users and complex tasks, including PPT generation, data analysis, and film and television production.

Doubao's official still emphasizes that basic services are free, but various signs indicate that the internal resource allocation of ByteDance is also changing.

According to LatePost, after some senior ByteDance executives visited Anthropic, ByteDance began to adjust its AI resource allocation, shifting the focus from mass - oriented products like Doubao to products serving enterprises.

Enterprise customers are willing to pay for scenarios such as model calls, office automation, customer service, marketing, and data analysis, and it is easier to justify high - cost computing power investment with efficiency improvements. For Volcengine, whether the Doubao large model can enter more enterprise scenarios is directly related to the revenue of the AI business.

Of course, unlike many independent model manufacturers, ByteDance is not short of money.

Douyin, e - commerce, advertising, and overseas businesses provide a stable cash flow. Previously, market rumors said that ByteDance's secondary transaction valuation reached $550 billion, a significant increase compared to the previous employee repurchase valuation.

For such a super unicorn, an IPO at the group level is not necessarily the top - priority choice.

However, it is obvious that ByteDance cannot accept Doubao having only traffic but no commercial breakthrough in the long run.

Different from the logic of the short - video content business in the Douyin era, the more active the users of the large model are, the clearer the inference cost is on the books.

Before the paid data is actually disclosed, it still has to answer a question: Can hundreds of millions of DAUs be converted into sustainable revenue?

As the testing of Doubao's professional version is about to be implemented, this question may soon have a preliminary answer, which also makes the discussions about Doubao's spin - off and external financing closer to reality.

2

If the driving factor for the spin - off is commercialization, the practical factor comes from the organizational level.

The real key signal comes from the incentive system. After the emergence of Doubao shares, ByteDance's AI business has taken an independent path at the organizational level.

Recently, the Seed team issued an internal notice that the latest price of Doubao shares has been adjusted to $14.85. Two months ago, ByteDance just initiated the first repurchase of Doubao shares at a price of $13.08.

The value of Doubao shares has been re - evaluated twice in just over two months. While the price continues to rise, it also widens the price gap between the AI business options and ByteDance's options.

This kind of incentive design has two practical effects.

First of all, it is of course to retain talent. The talent competition faced by large - model teams is becoming increasingly fierce, and it is not uncommon to hear rumors that top employees in the industry receive packages worth hundreds of millions. For core business members, only having group options can hardly fully reflect the growth rate of the model business and the scarcity of talent.

Another important effect is pricing. Virtual shares are not the same as external equity, but they can provide an internal valuation anchor for the team, the company, and potential investors. The continuous increase in the price of Doubao shares means that ByteDance has internally confirmed that the growth of the AI business is worthy of separate re - evaluation.

From this perspective, Doubao shares have completed the most important step before the spin - off: enabling the AI business to have a value expression different from that of the group options.

If the AI business continues to be fully incorporated into ByteDance's group valuation, the Doubao team can only share the overall growth of the group. Doubao shares separate the large - model business, which can not only retain model talent but also reserve a way for future external valuation.

External AI companies have pushed talent incentives to a higher level.

Take OpenAI as an example. The employee equity of this company is not traditional stock options.

Public information shows that in the early days, OpenAI issued profit participation units (PPU) to its employees. An employee's offer usually consists of two parts: cash compensation and PPU. PPU is not an ordinary stock, and its return is linked to OpenAI's future profit distribution and company value.

Anthropic provides another example. Public reports show that after Anthropic's valuation quickly rose from about $18 billion at the end of 2024 to nearly $350 billion, the value of its employees' options was quickly re - evaluated.

An engineer who joined at the end of 2024 was granted 60,000 options with an exercise price of $13. The value of the vested part has reached $4 million to $5 million, and the value of the fully - granted part may reach $18 million to $20 million.

Meanwhile, a number of established large companies have also been forced to follow suit. Previously, Meta offered multi - year, multi - million - dollar compensation packages to AI researchers to form a super - intelligent laboratory.

These cases point to the same change: the long - term income of AI talent is increasingly directly tied to the valuation of the AI business.

ByteDance's provision of Doubao shares to the Seed team is in line with the approach of these American competitors.

This independent options model has naturally made the income system of AI business employees highly tied to the AI business and separated from the group level.

However, Doubao shares are not external equity. They solve the problems of internal incentives and internal pricing. If Doubao really wants to be spun off and seek external financing, it depends on how willing ByteDance is to restructure its large AI landscape.

3

Based on recent information, ByteDance's AI business seems to be building the theoretical foundation for spinning off Doubao. However, when it comes to this company's AI landscape, it is not easy to separate this business.

For a long time, rumors about "ByteDance going public" have circulated in the industry from time to time, but there has never been any real action. For a super unicorn with a stable cash flow, an overall IPO has never been the top - priority choice.

However, the AI business seems to have changed this logic.

The investment in the large - model business is long - term and exponential. Model training, inference chips, data centers, AI Infra, model talent, and product subsidies all require continuous investment.

The more successful Doubao is, the more ByteDance needs to find ways to handle its revenue and capital. Outside the Chinese market, global large - model companies are also heading towards the capital market.

OpenAI and Anthropic have both entered the critical stage of preparing for an IPO. Obviously, large - model companies with high growth and high investment need to use the public market to handle long - term capital expenditure and talent incentives.

In addition, the public market still has a strong interest in AI companies, especially top - tier AI giants like ByteDance. Against this background, it makes sense from a capital perspective for ByteDance to spin off Doubao or a larger - scale AI business.

ByteDance can retain control while allowing the AI business to introduce external capital; it can stabilize the model team with an independent valuation; and it can partially separate the high investment in the AI business from the profit framework of mature businesses such as Douyin, e - commerce, and advertising.

Mature businesses focus on profits, while the AI business focuses on growth.

These two valuation logics in one group will naturally restrict each other. If Doubao's commercialization testing produces results, the imagination of external financing and independent listing will be further enhanced.

On the other hand, there is already a reference for business spin - off within ByteDance.

In February this year, market rumors said that Dongchedi was considering an IPO in Hong Kong and might raise $1 billion to $1.5 billion. Public information shows that Dongchedi completed a financing of about $600 million in 2024, with a valuation of about $3 billion.

However, the fact that Dongchedi can be discussed for an independent listing is related to its business boundaries.

Its business is highly vertical, focusing on automobile information, vehicle model data, transaction leads, dealer services, brand advertising, and value - added services. It can continue to cooperate with Douyin, or it can seek external financing and conduct independent accounting.

The landscape of Doubao, or ByteDance's AI business, is much more complex.

The Doubao App is just the front - end. Behind it are the Seed team, the Doubao large model, Volcengine, the AI Infra layer, and a host of other teams. Spinning off Doubao will affect not only the equity structure of an App.

The Seed team supports both Doubao and the AI transformation within Volcengine and ByteDance. The Doubao large model is sold externally through Volcengine and is also integrated into Douyin, CapCut, e - commerce, advertising, and enterprise services internally. If Doubao becomes independent, the model R & D cost, licensing revenue, and internal call price all need to be recalculated.

In addition, Doubao can get support from ByteDance's ecosystem thanks to the entrances and scenarios provided by Douyin, Toutiao, CapCut, Volcengine, etc. After the spin - off, this support will either continue in the form of group cooperation or become commercial procurement.

Both methods will bring problems.

If the group cooperation is too strong, investors will question Doubao's independence. If it takes the commercial procurement route, the pressure on Doubao's profit margin will be magnified. The more Doubao depends on the Douyin Group for traffic, data, and scenarios, the harder it will be to price it like an independent company.

This is also the biggest difference between Doubao and Keling.

Keling is an AI video business, with more concentrated product, revenue, and technology directions. Doubao covers general assistants, search, creation, office work, voice, video, enterprise services, and model output. The more complex the business landscape, the more difficult the spin - off.

More importantly, ByteDance also needs to consider strategic control.