When it comes to competitors, 10-year-old streaming service Curiosity Stream has more than its fair share in the factual entertainment category. From giants such as Netflix to niche players including MagellanTV, nonfiction content has no single source that dominates the way, say, Crunchyroll does in anime.
But what separates Curiosity Stream from the pack is something not even evident in the product itself, which has attracted more than 20 million subscribers worldwide.
Think of it as a superhero with a secret identity: By day, it’s built itself as a brand for natural history docudramas. But by night, the Maryland-based company is flexing a newfound superpower no other market entrant in its core streaming business has dared to try — moonlighting as a leading licensor of ethically sourced video and audio to AI companies ravenous for high-quality training data.
The company is mum on which AI giants are currently feeding on its massive library of 1 million-plus hours of content offered as multiple distinct datasets.
But over a year since striking its first deal, the results speak for themselves. After losing $51 million in 2022, Curiosity Stream is on the road to profitability, largely due to three consecutive quarters of growth for its AI licensing unit, which is powering third-quarter revenue guidance for the company of $15 million-$18 million.
That may be peanuts to the larger entertainment companies Curiosity Stream CEO Clint Stinchcomb believes will be too intimidated by the complexities of licensing to follow him, particularly because scripted content is freighted by far more restrictive rights.
“It’s certainly easier when you’re in the nonfiction space than when you are in the world of scripted entertainment and you have talent to deal with and guilds,” Stinchcomb said on the latest episode of the Luminate podcast In the Lab. “It’s fraught with lots of issues.”
By contrast, Curiosity Stream is less encumbered and actually benefits from its singular focus on factual fare, such as nature documentaries that can provide video model developers with hours of high-quality footage of landscapes or animals.
Curiosity Stream’s training trove is a mix of owned content and licensed fare from third parties who are happy to do a revenue share rather than navigate an unproven new business that can be a distraction to media companies that already have enough challenges on their hands.
To crack the AI licensing market, having video at scale to sell is only half the battle; properly handling the metadata attached to that video is like learning to speak a new language for film and TV producers.
“If you can’t label and annotate at scale and process raw video, things that heretofore have been the province of technology companies — it’s a much harder proposition,” said Stinchcomb.
Best case scenario: Curiosity Stream is a first mover learning the ropes of a licensing business where a head start will make all the difference in securing revenue from AI partners. Worst case: Those partners’ training needs unexpectedly change in a nascent field, and AI licensing ends up a blip on the company’s balance sheet in the long term.
The ultimate longevity of the licensing market for AI training data most notably hinges on how courts rule on AI copyright cases where the fair-use defense is the biggest question. Nevertheless, initial results raise profound questions about what the future of content-monetization strategy looks like across the industry.
Stinchcomb remains optimistic. “Over this past century, value creation in media consistently migrates to those who can capitalize on paradigm-shifting innovation,” he said. “I think we’re at the threshold of the most profound advancement yet.”
