Curiosity Stream Pivots: AI Licensing Expected to Drive Majority of Revenue by 2027
Curiosity Stream is tired of fighting the streaming wars. Its new plan? Selling the ammo. The documentary service, founded by Discovery Channel creator John Hendricks, is betting the farm on a radical pivot: by 2027, it expects selling data to AI companies will generate more than half its total revenue. It’s a stark departure from chasing subscribers in a saturated market, effectively repositioning the company from a consumer streamer to a backend data arms dealer.
The forecast dropped during the company’s Q3 2025 earnings call on November 18, where executives outlined how their archive of factual content is being repackaged for the insatiable maw of AI model training. With competitors fighting over churn rates, Curiosity Stream is looking upstream, banking on the desperate need for "clean," verified video data from giants like OpenAI and Anthropic.
The Math: Licensing Pays Better Than Subscriptions
The financial logic here is blunt: high-margin B2B licensing beats the grind of acquiring individual subscribers. Curiosity Stream projects total annual revenue will hit between $100 million and $150 million by 2027, with AI licensing deals contributing a massive $50 million to $75 million of that pie.
That is an aggressive ramp-up. Right now, AI deals account for roughly 10-15% of the company's 2025 revenue. Barely a year ago, that number was a rounding error—under 5%. The company is effectively wagering on a 3x increase in AI revenue share over the next 24 months.
Investors, at least, seem to like the math. Q3 2025 saw a 200% year-over-year explosion in licensing revenue, pushing total quarterly earnings to $15.2 million—a 15% bump over last year. The strategy is also stopping the bleeding; net losses narrowed to $2.5 million, down significantly from the $4.1 million lost in Q3 2024.
Solving the "Ground Truth" Problem
Why would tech behemoths care about a relatively small library of nature docs? As generative AI pivots from text to video, hallucinations—AI just making things up—are a massive liability. Models need "ground truth" to understand how a cheetah actually runs or how a volcano erupts. Unlike the messy, copyright-trapped sludge scraped from the open web, Curiosity Stream offers a sanitized, legally safe library of over 3,000 hours of premium content.
During the call, CEO Jonathan Huberman pitched this library as "uniquely suited for AI training." The theory is that factual narration paired with verified visuals helps models grasp real-world physics and facts better than random YouTube clips. To grease the wheels, the company launched an "AI-Ready Content Portal" this quarter, allowing developers to grab metadata-tagged clips without the technical headaches associated with scraping.
However, there is a valid question of scale. In the context of Large Language Models (LLMs) and video generators trained on millions of hours of footage, 3,000 hours is a drop in the bucket. Curiosity Stream isn't offering volume; they are offering purity. The bet relies on the idea that AI developers will pay a premium for a specialized, high-fidelity "finishing school" for their models, rather than just raw bulk data.
Market Reaction and the Archive Gold Rush
Wall Street is buying the narrative. Curiosity Stream's stock ($CURI) jumped 7% following the earnings release, trading around $1.50 with an $80 million market cap. Analysts are eyeing those 80-90% margins on licensing deals—a dream scenario compared to the razor-thin 40% margins of traditional streaming.
This move mirrors a wider industry scramble. Everyone from Reuters to Getty Images is trying to monetize their archives before AI simply learns to generate content without them. But while others sell text and photos, Curiosity Stream is trying to corner the market on factual video.
Crucially, the company insists this is a content play, not a privacy breach. Huberman stated explicitly that subscriber data is off the table; the deals are purely about the video library. The legacy streaming service isn't dead—it still holds a stable 25 million subscribers—but with growth there flatlining, it's clear where the company’s attention has shifted.
As the race for training data intensifies, Curiosity Stream has shown its hand. The value of its library isn't just in entertaining humans anymore—it’s in teaching machines how the world works.