Good morning!
The Medium delivers in-depth analyses of the media marketplace’s transformation as creators, tech companies and 10 million emerging advertisers revolutionize the business models for “premium content”.
[Author's Note: My monthly Medium Shift opinion column —”The Media Revolution Will Be Prompted”— went live on Tuesday. I wrote about how the AI tools that create videos from text prompts may threaten creators, and why that leaves Netflix vulnerable after its recent pivot to an ad-supported model.]
YouTube’s next chapter with generative AI presents existential questions about the value of content libraries both within and beyond its ecosystem. In the U.S. alone, YouTube has now secured the top spot among Nielsen’s The Gauge—its monthly snapshot of viewing across linear TV and streaming on U.S. TV screens— for 12 consecutive months. Everyone is competing with YouTube’s enormous and constantly expanding library: More than 500 hours of video are uploaded to YouTube every minute.
In my latest Medium Shift column for The Information— “The Media Revolution Will Be Prompted”—I wrote about creators’ and others’ fear of the “flooding” platforms with AI-generated content because it sets a lower bar. Some models—like that of creator Jordi Maxim van den Bussche, aka Kwebbelkop—have lowered the barriers to entry for content creation so that *so* many videos to be produced and uploaded so quickly that YouTube could easily be “flooded” with AI-generated content.
There was a second—perhaps larger—concern expressed by YouTube creators Colin Rosenblum and Samir Chaudry that YouTube audiences may develop an “appetite” for inauthentic, AI-generated storytelling at the expense of human-created content. That is implicitly a fear about how algorithms on ad-supported platforms like YouTube and Spotify tend to serve content that is both brand-safe and generates the most impressions for the platform.
Two questions that this combination of generative AI tools like OpenAI’s Sora and a consumer appetite for its output are:
If AI-generated content becomes considered brand-safe, why wouldn’t YouTube’s algorithm promote it to users to drive more engagement? And,
In response, how will everyone else invest to compete?
As Vulture’s Joe Adalian wrote last August, the rest of the competition is facing a generation decline in demand for their libraries: “At some point, platforms are going to run out of quirky linear shows from the 2000s and 2010s that zoomers [Generation Z] can rediscover on streaming.” With limited resources and limited options, all streamers—including Netflix—are all going to need to rely on YouTube’s algorithm as their “frenemy”.
Key Takeaway
As generative AI emerges as a competitive threat to streaming services, YouTube is being asked to take on the role of an imperfect, self-interest bulwark to protect the entire subscription streaming ecosystem from a problem that it has mixed incentives to regulate.
Total words: 1,600
Total time reading: 6 minutes
The [Insert Platform Here] Paradox
I wrote about this dynamic before in “AI Is YouTube's Frenemy”, but focused on how “YouTube’s challenge is ultimately that it is well-positioned to harness the efficiencies of AI for content creation. AI-savvy creators whom it relies upon to drive and grow engagement on its platform are also well-positioned.” This new “frenemy” dynamic—which has become more tangible with the emergence of tools like Sora— reflects the open question of how YouTube’s algorithm will navigate and evolve with the inevitable influx of generative AI content to come.
But, there is already a “frenemy” dynamic in place between streamers and YouTube’s algorithm as I argued in “‘The Office’ Is Succeeding On YouTube, Less So on Peacock”. I identified a problem that could be labeled “The Netflix Paradox”, “'The Office' Paradox” or “The YouTube Paradox”:
Streaming services have invested billions into exclusive intellectual property libraries at a time when the value of IP is fragmenting across platforms;
The best business models for monetizing this IP should be PARQOR Hypothesis businesses because they centralize the value of IP and monetize it in multiple ways; but,
Without a centralized model for the IP, the YouTube ecosystem and algorithm may be more valuable to building fan bases for IP than the exclusivity of a “walled garden”.
The core argument was that YouTube had disintermediated the relationship between the owners of valuable IP and their fan bases. I used the example of Peacock’s “The Office”: New generations of viewers consume “The Office” on YouTube (and TikTok) and less so on Peacock. Clips and memes generate more word of mouth and engagement for the show outside of Peacock than the full episodes or the “super fan” episodes of The Office offered exclusively on Peacock. We know this because “The Office” was Nielsen’s most streamed show in the U.S. in 2020, and it has not ranked in the Top 10 since.
NBCUniversal signed a five-year, $500MM investment in “The Office” in 2020. Four years later, Peacock lost $2.8 billion in 2023, so it is hard to imagine that “The Office” has had the economic value to the platform that NBCUniversal assumed it did back in 2020. One only has to look at the success of “Suits” to understand why: After Netflix licensed it non-exclusively from NBCUniversal in June, it became the most-streamed series in the U.S. across all platforms for 2023. It was all but unwatched on Peacock relative to its consumption on Netflix.
From the perspective of both NBCUniversal and the rights holders of “The Office”, they are all best off with their content being distributed on Netflix. But, even if that solution makes economic sense, they all will remain powerless against both disintermediation by YouTube and competition from an influx of supply and demand of generative AI content.
Netflix & “What We Watched”
These new competitive dynamics with generative AI raise a difficult question for both NBCUniversal and its Peacock platform, and legacy streamers more broadly: For how much longer can they rely on once-popular TV titles like “The Office” to drive subscription growth and also reduce churn?
Libraries were considered a sunk cost and streaming was considered to be the next wave of “free money” for licensing those libraries when Amazon Prime Video (2006), Netflix (2007) and Hulu (2008) launched almost 20 years ago. Almost 20 years later, there are more streaming services on the marketplace and—excluding their library licensing models with third parties—none except for Netflix and Hulu are profitable. Streaming subscription models have become a costly source of recurring revenues.
Adalian also highlighted how in recent decades, the total episode run of a series has shrunk, too: “200 episodes of a show, that meant bigger syndication paydays, more DVDs to sell, and more time to sell to advertisers — all of which turned hits into ATMs for all involved.” After Netflix decided to create exclusives for its platforms—all of which average less than three seasons in length—and other legacy media streamers followed suit, all streamers elected to pursue a similar model of shorter seasons.
Netflix’s release of “What We Watched” highlighted that its investment rationale for content actively avoids this level of scale. Rather, it is “all about whether a movie or TV show thrilled its audience — and the size of that audience relative to the economics of the title.” As I noted in “Market-Making Beyond Netflix's Walled Garden”, the extraordinary success of TV series like “The Night Agent” within the Netflix ecosystem reflects how the “efficient economics and savvy risk-taking within the walls of Netflix are only possible with Netflix's "plumbing" behind "the poetry" of “on-demand, personalized, and available on any screen.”
In other words, even the market leader no longer sees an economic rationale in the type of production model that may compete with an influx of supply and demand of AI-generated content on YouTube, alone.
A Dynamic Problem
Adalian noted that for streamers to return to producing as many as 80 or more episodes “would likely require some other big paradigm shifts, such as getting rid of the cost-plus model under which streamers pay the full cost of production for a show plus a bonus to studios.” Perhaps, but such a solution would still hit the same logical endpoint: the best outlet for streaming content remains Netflix.
Overall, Netflix is increasingly financing series and movies—$17 billion in spend planned for 2024—that balance the ability to “thrill” large enough audiences to justify the economics. Games are much cheaper than an 80-episode series: Netflix paid $72 million for a Finnish game studio in 2022 or 14% of what NBCUniversal paid for “The Office”. Games also offer a differentiated value proposition relative to Generative AI. But, Netflix will also be the first to tell us that it competes for *attention* with other platforms.
Another solution lies within Netflix’s bet on games. As I argued in “Market-Making Beyond Netflix's Walled Garden”, “personalization algorithms ultimately connect supply to demand and then evolve to increasingly bet on what the most scalable forms of “connection” may be.” To date, that content has been streaming, but now games seem to be an increasingly important element of those connections, too. The problem is that, for now, the demand seems low. According to data from Sensor Tower, games released by Netflix were downloaded 81.2 million times from Apple's App Store and Google Play in 2023. 53% of that total came in the fourth quarter alone, likely due to the release of the Grand Theft Auto trilogy in November.
The competitive reality is that the better YouTube is at capturing attention with both creator and generative AI content, the harder it will be for Netflix and other services to win and keep streaming subscribers. That puts YouTube in a powerful position: It already undermines the subscription models of its legacy media competition by simply being a hub where clips and memes can go viral, despite being from shows paywalled elsewhere. Once its algorithm pushes more AI-generated content, it can undermine the subscription model of Netflix and other services simply by capturing more attention.
So, the ultimate danger of “flooding” is that it puts YouTube’s algorithms in the position of needing to protect the YouTube and third-party ecosystems but to also favor the content that is both brand-safe and generates the most impressions for the YouTube platform. It is being asked to take on the role of an imperfect, self-interest bulwark to protect the entire subscription streaming ecosystem from a problem that it has mixed incentives to regulate.