Kling's performance forecast for the third quarter (Q3) is rather lackluster.

In Q3, Kling's revenue amounted to 850 million yuan, showing negligible growth from the 800 million yuan recorded in Q2.

More subtly, in its latest earnings report, Kling's management refrained from disclosing the Annual Recurring Revenue (ARR), only stating an estimated annual loss of 5-6 billion yuan. This is particularly concerning given that Kuaishou has already invested 26 billion yuan in AI capital expenditure by 2026.

Despite substantial investments, revenue growth remains stagnant.

As a result, Kuaishou's stock price plummeted by 11 points today. Year-to-date, Kuaishou's shares have fallen nearly 50%.

This downturn is largely attributable to ByteDance. According to LatePost, in June this year, ByteDance's Seedance video generation model's ARR peaked at 2 billion US dollars, approximately 14.3 billion yuan, with monthly revenue exceeding 1 billion yuan.

The short-term demand for video models is limited. While ByteDance reaps substantial profits, its competitors can only watch helplessly.

Of course, Kuaishou is not the only company affected by ByteDance's competitive edge.

Among investors in Hong Kong stocks, a prevailing viewpoint has emerged: since ByteDance is not publicly listed, the best way to invest in ByteDance's AI business is to short the Hang Seng Tech Index.

The rationale is straightforward. Many of the weighted stocks in the Hang Seng Tech Index, such as Tencent and Alibaba, are direct competitors of ByteDance in the AI arena.

As ByteDance's AI business strengthens, this viewpoint gains more traction. According to the latest disclosed data, ByteDance's large model business ARR has reached 4 billion US dollars, surpassing the combined ARR of all other domestic model companies.

Thus, the Hang Seng Tech Index's decline over the past year is closely linked to ByteDance's ascent. Internet investors, repeatedly impacted by ByteDance, have gradually formulated an investment rule:

You don't have to be time's ally, but at least avoid being ByteDance's adversary.

/ 01 / Models Are Not Kling's Primary Weakness

Many attribute Kling's inability to compete with Seedance to its slightly inferior model capabilities. However, this perspective underestimates ByteDance.

ByteDance's true strength lies in its creation of an unprecedented commercial closed loop in the AI video space, with AI short dramas at its core.

Short dramas are the primary consumers of video models.

Producing a 60-episode, 2-minute-per-episode live-action short drama costs 400,000-600,000 yuan. An AI digital human short drama with a similar feel costs 100,000-200,000 yuan, with some projects even cheaper, and the production cycle is significantly shortened. For most short drama companies, transitioning to AI short dramas is a logical choice.

Previously, ByteDance's advantage in AI short dramas was mainly in distribution channels.

According to QuestMobile data, as of May 2026, the monthly active users of all independent short drama apps were approximately 402 million, with Hongguo Free Short Dramas alone accounting for 356 million. In other words, ByteDance has captured 90% of the domestic short drama market's traffic.

According to Nomura's research data, Hongguo has secured over 60% of the AI short drama market share.

Compared to its near-monopolistic advantage in distribution channels, ByteDance took a more circuitous route in video models, not achieving State-of-the-Art (SOTA) performance until the end of 2025 and then dominating the competition.

Consequently, ByteDance now possesses both the strongest video model and the most efficient content distribution channel, giving rise to a new business model.

ByteDance sells Seedance's API to short drama companies, which then distribute their content on Douyin and Hongguo. After users start watching, Hongguo and Douyin convert these viewing hours into new ad inventory and sell it to advertisers, completing a flawless closed loop.

In essence, the same AI short drama can sequentially generate model revenue, traffic revenue, and ad revenue within ByteDance's ecosystem.

More importantly, a significant portion of the production costs saved by AI ends up back in ByteDance's pockets.

In the AI short drama business model, the primary expenses are computing power and traffic acquisition.

Traffic acquisition has always been the most substantial expense for traditional paid short dramas (IAP). Typically, 80%-90% of revenue is spent on user acquisition. In other words, for every 100 yuan in recharge revenue, 80-90 yuan may be spent on acquiring users.

AI short dramas are no exception. In June 2026, Caijing.com quoted an AI short drama practitioner as saying that current AI short drama traffic acquisition costs account for about 70% of total revenue.

In other words, most of the production costs saved by AI become Seedance's API bills. Seventy percent of total revenue becomes ByteDance's traffic acquisition fees. It all flows into the same pocket.

Previously, I thought short dramas were already an excellent business model. I didn't anticipate that Hongguo Free + Seedance could elevate this model to an even more profitable level.

With this cycle, Seedance operates on a completely different level compared to its competitors in the video model space.

The model belongs to ByteDance, the traffic belongs to ByteDance, and the ads belong to ByteDance. How much content creators earn is never in their own hands; it entirely depends on ByteDance's discretion.

Thus, the ecological niches in the entire AI short drama industry chain ultimately boil down to: ByteDance and everyone else.

/ 02 / AI Short Dramas Have Reached a Short-Term Peak

The previous surge in video model revenue was driven by the explosive growth of AI short dramas.

In the first five months of 2026, the domestic AI drama and comic market size reached 22 billion yuan. In the overall short drama market, AI short drama traffic acquisition accounted for 70%-80% of the total, with live-action short dramas accounting for only about 30%.

While the significant increase in AI short dramas has driven growth in video model revenue, it has also introduced another issue: most producers are no longer profitable.

The success rate of a high-quality live-action short drama going viral can reach 10-20%, while the success rate for AI short dramas is only 2-6%. According to Celia, head of Shenzhen Chuangliang Market, the hit rate (over 100 million views) for AI live-action comic dramas is even less than 1%.

Moreover, AI short dramas have a relatively short popularity cycle, generally lasting only 2-4 weeks.

As if the increasing competition in content wasn't enough, platforms are also reluctant to fully support AI short dramas. The reason is simple: allocating traffic to AI short dramas is not cost-effective.

The vast majority of Hongguo's revenue comes from advertising. According to Nomura's research, in the first half of 2026, Hongguo allocated 30% of its traffic to AI short dramas, resulting in 2.8 billion yuan in ad revenue from AI short dramas, while Hongguo's total ad revenue was 19.3 billion yuan.

In other words, AI short dramas used 30% of the platform's traffic resources but only generated 15% of the revenue. The reason is straightforward: brand advertisers are less willing to place ads on AI short dramas and prefer live-action short dramas.

At the same time, if the platform over-promotes AI content, it will compress users' overall viewing time, cause user aesthetic fatigue, reduce user retention, and ultimately impact ad monetization efficiency. Ad revenue is directly tied to user dwell time and user content emotional attachment.

So, for Hongguo, most AI short dramas are low-quality inventory: large in quantity, low in value, and detrimental to the platform.

Because of this, Hongguo has begun to reduce the supply of AI short dramas.

On the one hand, it adjusted the revenue-sharing mechanism. In May this year, industry insiders revealed that Hongguo Short Dramas canceled the guaranteed revenue for AI live-action scripts. All scripts reviewed after April 27 would have no guaranteed revenue and would instead operate on a pure revenue-sharing basis, with a sharing ratio of 20%.

On the other hand, Hongguo is also controlling the proportion of traffic allocated to AI short dramas, maintaining it at 30%.

The former has exacerbated the unprofitability of AI short dramas, while the latter has firmly capped the traffic growth of AI short dramas. With both factors combined, the supply of AI short dramas will shrink significantly, and a short-term peak is inevitable.

As the largest application scenario for video models, when the growth story of AI short dramas comes to a phased end, it naturally affects Kling's performance growth.

/ 03 / Conclusion

ByteDance's triumph in video models is merely a microcosm of its broader success in AI.

Currently, ByteDance's large model business ARR has reached 4 billion US dollars. What does that signify? It exceeds the combined ARR of all other domestic model companies.

Other significant players include Alibaba with 1.5 billion US dollars, Zhipu with 1 billion US dollars, and DeepSeek with around 400 million US dollars. Together, they are no match for ByteDance alone.

In primary market circles, there's a saying: Chinese venture capitalists (VCs) are divided into two groups: those who invested in ByteDance and those who didn't.

Now, secondary market investors are also categorizing internet companies into two groups: those that compete with ByteDance's main business and those that don't.

This may be an exaggeration. But one thing is certain: for investors, avoiding competition with ByteDance's core sectors is advisable.

Don't believe it? Just look at Kuaishou.

By Xiaobai