99% AI, $500M Run Rate: Pocket FM Just Became the Case Study for AI-Native Media

For two years the media industry has argued about whether generative AI can build a sustainable content business rather than just cheapen an existing one. Pocket FM, the Bengaluru- and Culver City-based audio storytelling platform, has just handed the debate its cleanest data point yet. In an interview published September 10, 2026, co-founder and CEO Rohan Nayak told TechCrunch that the company has doubled its annualized revenue run rate to $500 million over the past year — while artificial intelligence now powers 93% of its overall catalog and is used to produce 99% of its new content.

The numbers behind that transition are startling. Content production costs have fallen roughly 80x. One hundred hours of content, which previously took about a year to produce, can now be made in a single day. The platform’s more than 550,000 creators now generate about 2.5 million hours of AI-powered content a year — a figure worth setting against the startup’s entire catalog of roughly 100,000 hours just two years ago. The library now spans more than 770,000 audio series.

Humans stay in the loop — on purpose

What makes Pocket FM’s model notable is not the volume of AI output but where the company deliberately kept humans. Nayak was emphatic that the platform still relies on human creators for ideas and storytelling, with AI turning those concepts into finished audio at scale. “We want to create great IPs that last 100 years, and that needs humans,” he told TechCrunch.

That philosophy is embodied in AI head Vasu Sharma, a former Meta and Tesla scientist, whose team has trained Pocket FM’s own in-house models for creative writing and text-to-speech — trained not on the open internet, but on years of proprietary production data and, critically, on signals showing how listeners actually engage with stories: where they pay, where they binge, where they drop off. The flywheel is the asset: every listening session teaches the models what makes an audience stay.

The strategy appears to be working on the retention side too. Pocket FM’s 12-month revenue retention rate has climbed to 76%, up from 44% two years ago, which Nayak attributes partly to sheer catalog depth — with more stories available, the platform can match a far wider range of listener preferences and keep paying users from churning.

Follow the money: $415M from episodes, $85M from ads

The revenue mechanics are unusually transparent for a private company. Of the $500 million annualized run rate, roughly $415 million comes from users paying to unlock individual episodes — a microtransaction model borrowed from gaming — while about $85 million is advertising. The run-rate figure is calculated by multiplying monthly revenue by 12, Nayak explained, rather than from contracted recurring revenue.

The trajectory itself tells the story of the AI ramp: about $250 million a year ago, $430 million by April 2026, and $500 million now. Growth drivers included the AI content push, expansion into the U.K., Germany, and France, and the launch of user-generated content in the U.S. That U.S. market — which grew around 70% over the past year — now accounts for roughly 70% of the entire run rate, despite Pocket FM beginning life as an India-first platform. The platform claims more than 250 million listeners across over 20 countries.

Individual titles are performing like small businesses: 96 series have generated more than $1 million each in revenue, and 13 have crossed $10 million.

Pocket Saga: the fully AI experiment

The most strategically revealing detail is Pocket Saga, a microdrama app launched just three months ago that has already reached an annualized revenue run rate of about $15 million. Unlike Pocket FM, where humans remain involved in story development, Pocket Saga’s content is entirely AI-produced — the company even repurposes successful Pocket FM audio stories into AI-generated video microdramas, with no traditional live-action production involved.

The app is currently U.S.-only, and it functions as a live experiment in how far the “no humans in the creative loop” model can be pushed commercially. Parent company Pocket Entertainment plans to expand into at least two more entertainment formats over the next five years, and aims to license successful stories into books, television, and film.

A $2 billion raise in the works

Alongside the growth disclosure, Pocket Entertainment confirmed it is in talks with investors about raising fresh capital — reported at $100–120 million at a valuation of around $2 billion. Nayak did not deny the talks but said the company is under no immediate pressure to raise, declining to specify amounts. He noted that a new round would primarily fund further AI investment and the new entertainment formats. The company says it is profitable and generating positive cash flow on an adjusted basis, though it declined to disclose margins. An IPO is not expected within the next 24 months, with both Indian and U.S. listings kept on the table.

Why this matters beyond one startup

The debate over AI and creative work has mostly run on anecdotes — a slop-filled stock image here, a displaced voice actor there. Pocket FM offers something rarer: a full P&L-style view of what happens when a media company re-architects itself around AI production end to end. The results cut both ways. On one hand, the 80x cost collapse and 2.5 million annual content hours vindicate the bull case that AI radically expands what a content business can supply. On the other, the company’s own insistence that humans remain essential for “IPs that last 100 years” — and its 76% retention being credited to matching human-crafted story variety — suggests the winning formula is human taste with AI manufacturing, not AI alone.

The uncomfortable question the industry will now have to answer is what happens to the economics of audio drama competitors, audiobook publishers, and scripted podcast studios that don’t own a proprietary engagement dataset to train their own models on. Pocket FM’s moat isn’t a foundation model — it’s eight years of listener behavior data wrapped around one. For everyone else, the 80x cost curve is coming whether they trained for it or not.