Kling AI passes US$100 million annualized revenue

Background

Kling AI launched in June 2024 ^[1]. It reaches users through two channels that both appear in the figures below: monthly subscriptions bought by individuals, and an API that other companies and developers call from their own software. A subscription turns a user into recurring monthly revenue; an API turns usage into metered revenue billed to businesses rather than individuals.

What happened

On the product's first anniversary, Kuaishou said that Kling AI's annualized revenue run rate had passed US$100 million in March 2025, the tenth month after launch ^[1]. The same announcement reported that monthly subscription bookings exceeded RMB 100 million in April and May 2025 ^[1].

The run-rate figure needs explaining, because it is easy to misread. Kuaishou defines annualized revenue run rate as the current month's revenue multiplied by 12 ^[1]. It is not a forecast, and it is not a trailing twelve-month total: it takes one month — March 2025 in this case — and asks what a year would look like if that month repeated twelve times. The number therefore moves with the latest month, and one strong or weak month moves it sharply. Nor does it account for costs, so it measures scale rather than profit.

The subscription number is a different kind of measure. Monthly subscription bookings count the subscription orders placed in a given month; April and May 2025 each came in above RMB 100 million ^[1]. Bookings and recognised revenue are not the same thing — one records what customers committed to, the other what the company has earned as the service is delivered — so the two disclosures describe the same business from different angles. They are also stated in different currencies, one in US dollars and one in renminbi.

Key details

Kling AI went through more than 20 iterations in its first year ^[1]. That pace carried into the anniversary period. Kling AI 2.0 rolled out to global users in April 2025 ^[1]. At the end of May 2025, Kling AI 2.1 debuted in two modes — Standard at 720p and High Quality at 1080p — and a premium 2.1 Master Edition was released alongside it ^[1]. Read together, the version history and the resolution tiers describe a product being pushed on two fronts at once: 2.0 widened access by reaching global users, while 2.1 added a higher-resolution option and a top-end edition above it. Offering several tiers is a pricing decision as much as a technical one, since each tier gives customers a different price point to buy at.

On the business side, the API had served more than 10,000 corporate clients and developers ^[1]. API customers and subscribers are counted differently and pay differently: a developer embedding the model in an app generates usage that rises and falls with that app, while a subscriber pays a recurring fee for as long as the subscription lasts. Holding both channels means the revenue base is not tied to a single customer type or a single way of paying.

Why it matters

Revenue figures for individual AI products are disclosed rarely enough that each one is worth reading carefully. This one gives an outside observer a concrete data point on whether a subscription-plus-API combination can produce a nine-figure annualized run rate within a year of launch, and at what point in the product's life that happens: the US$100 million mark was crossed in the tenth month, with monthly subscription bookings above RMB 100 million arriving in the two months after that ^[1].

The caveats matter as much as the headline. A run rate is one month multiplied by twelve, so it can fall as quickly as it rose, and the next month's report can tell a different story. Bookings are not recognised revenue. A client count of more than 10,000 says nothing about how much each client spends. And an anniversary announcement is a milestone chosen for communication as much as for measurement. What the figures do establish is narrower and more solid: within its first year, Kling AI built a subscription business and an API business of a size its parent company was willing to publish — and it kept shipping new versions while doing so.