Pocket FM's $500M Run Rate: AI Makes the Episodes, Listeners Pay to Unlock Them

A Year of Production, Now Done in a Day

Pocket FM is an Indian company that began in 2018 as a platform for serialized audio stories. Co-founder and CEO Rohan Nayak told TechCrunch on September 10 that its annualized revenue run rate has reached $500 million. A year earlier it was about $250 million, and in April it was $430 million.

According to Nayak, AI has been used across 93% of the full catalog, and 99% of new content is now produced with AI. Production costs have fallen to roughly one-eightieth. One hundred hours of content used to take close to a year to make; now it takes a day, he said.

People have not dropped out of the process. Creators still come up with the ideas and shape how stories unfold, and AI turns those concepts into finished content. Vasu Sharma, the company’s head of AI and a former Meta and Tesla scientist, said Pocket FM trained its own models for creative writing and text-to-speech. The training material is years of production data plus signals on how listeners responded to stories.

Two Years Ago the Entire Catalog Was 100,000 Hours

The shift in volume is bigger still. More than 550,000 creators produce about 2.5 million hours of AI-assisted content a year. Two years ago, the platform’s entire catalog came to roughly 100,000 hours. The library now holds more than 770,000 audio series.

The company believes the extra volume kept listeners around. Twelve-month revenue retention rose to 76% from 44% two years ago, and Nayak cited a larger pool of stories matching different listener tastes as one reason. Ninety-six titles have each brought in more than $1 million, and 13 of those have passed $10 million.

Pocket FM also broke down where the money comes from. About $85 million of the run rate is advertising. The remaining roughly $415 million is what listeners pay to unlock episodes one at a time. The U.S. is the largest market: it grew about 70% over the past year and now accounts for roughly 70% of the run rate. Listeners number more than 250 million across more than 20 countries.

Wait for the Next Episode, or Pay

Mint explained the pricing in an April report. Pocket FM moved from a subscription-led model to freemium, mixing pay-per-episode purchases with premium plans. Listeners can wait for episodes to unlock over time, or buy coins in the app and listen right away.

Korean webtoon and web novel readers know this setup well. Kakao Page introduced its “gidarimyeon mureo” model, free if you wait, on October 23, 2014. Kakao says the model borrowed from the heart-refill mechanic of the mobile game Anipang: without paying, the next installment opens after a set time, and readers who do not want to wait pay. Of Pocket FM’s $500 million run rate, $415 million comes from the listeners who choose not to wait.

Monthly Revenue Times 12, and the Figures in the Books

“Run rate” is a term worth reading on its own. Nayak told TechCrunch the figure is monthly revenue multiplied by 12, not contracted recurring revenue. In an April LinkedIn post, he wrote that annual recurring revenue (ARR) had passed $400 million. The Mint article that reported the post does not say whether that number was calculated the same way.

Booked revenue moved at a different speed. According to Mint, Pocket FM’s revenue for fiscal 2025 was ₹17.68 billion (₹1,768 crore), up about 68% from ₹10.52 billion (₹1,051.97 crore) the year before. Its net loss narrowed from ₹2.08 billion to ₹1.65 billion. In the April LinkedIn post, Nayak said free cash flow had turned positive and the EBITDA margin was about 5%. In September the company said it is profitable and cash flow positive on an adjusted basis, but it did not disclose profit or margin figures.

Fundraising has not gone smoothly either. Mint reported on April 30 that Pocket FM had revived talks to raise $100 million to $120 million at a valuation of $1.5 billion to $2 billion. An attempt about nine months earlier failed to close amid investor concerns about backing a loss-making business, the report said. The previous round was a $103 million Series D in March 2024 led by Lightspeed Venture Partners and StepStone Group, at a valuation of about $750 million. Tencent and Naver are among its investors. In the September interview, Nayak did not deny the talks but said the company is under no immediate pressure to raise.

Korea’s AI Basic Act Separates Companies That Use AI From Those That Supply It

A company running the same business in Korea would start with the AI Basic Act, which took full effect on January 22, 2026. Article 31 requires businesses offering products or services that use generative AI to tell users in advance that the service runs on AI, and to label outputs as made by generative AI. Deepfake outputs that are hard to tell apart from reality must be disclosed in a way users can clearly recognize.

Who carries that duty is set out in the transparency guideline the Ministry of Science and ICT released ahead of the law taking effect. The obligation falls on AI businesses that provide AI products or services directly to users. Anyone using AI as a tool for work or creative production is treated as a user and left out of the obligation. The guideline’s example is a studio that makes and distributes a film with a video-generation model: that studio is a user, not a provider of an AI product or service.

Measured against that standard, a Pocket FM-style business splits in two places. Selling series the company finished with AI to listeners is close to the film studio example. But if the platform offers the AI production tools its creators use as a service, it could be read as a business providing an AI service directly to those creators. A Korean team that wants to bundle creator-facing AI tools and content sales in one app should assess the two parts separately.

Labeling methods differ by content type. In the guideline examples summarized by Korean law firm Shin & Kim, audio output does not need to be labeled for its whole running time; a spoken notice at the start saying it was generated with AI is enough. Output provided only inside the service can be flagged through on-screen notices, but output that leaves the service through downloads or sharing must carry the label in the output itself. The ministry has said it will run a grace period of at least one year, deferring fact-finding investigations and fines during that time.

Who Owns a Story Is Decided in the Contract

As more production runs through AI, the first questions in investor due diligence change. Rashi Bharadwaj, a partner at Indian law firm CMS IndusLaw quoted by Mint, named ownership and control of intellectual property in the platform’s content as the key legal issue investors would examine. Checking that requires reviewing the underlying agreements to see who holds the rights, along with their scope, exclusivity and duration, she said.

Pocket FM is also pursuing licensing deals that carry successful stories into books, TV and film. For a work a creator conceived and the company’s models finished, the terms of such a licensing deal depend on how the contract records those rights. A team preparing a creator-driven AI content platform is better off stating in its very first creator agreement who owns output that AI helped produce.

Parent company Pocket Entertainment is already taking the approach beyond audio. Pocket Saga, a microdrama app launched three months ago only in the U.S., makes all of its content with AI and turns successful Pocket FM audio stories into AI video without live-action filming. The app’s annualized revenue run rate is about $15 million.