Good afternoon!
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”.
Is the emergence of generative AI text-to-video platforms analogous to video technologies like Betamax and VHS that disrupted Hollywood in the 1970s and 1980s?
The argument was made in passing by Edward Saatchi—the founder of the generative AI platform Showrunner—whom I interviewed for my upcoming Medium Shift column for The Information.
His point has implied layers. VHS and Betamax disrupted theatergoing, so there is the technology angle. There is a legal angle, too: The film industry did not react well to these technologies—arguing copyright laws prevented the video recording capabilities of video cassette recorders (VCR)—and took their concerns to Congress and the Supreme Court. There was a cultural fear from filmmakers like Stephen Spielberg that home viewing of movies “would take the magic out of movies.”
History repeats itself in generative AI: New technological, legal and cultural fears have emerged that algorithmically-generated storytelling will disrupt moviegoing, television and streaming. Legal threats loom for learning language models trained on publicly available videos posted to YouTube and elsewhere.
But, the comparison falls apart between the business models. The retail cost of a VHS tape sold directly to consumers was around $100 in the late 1970s and early 1980s. Also, VCRs already cost consumers over $1,000 per unit in the 1980s. According to an inflation calculator offered by the Federal Reserve Bank of Minnesota, a videocassette in 1977 is now worth $482 in 2024, and a VCR $4,820. When video rental stores emerged in the mid-1980s—an “unintended consequence” of studios expanding their supply of videocassettes at $100—the result was over $2 billion annually in revenues in the late 1980s ($5.3 billion in 2024).
Today, anyone can access OpenAI’s Sora text-to-video model at a cost of $0.01-$0.10 per second of generated video. Bloomberg Intelligence projects the generative AI market will grow to $1.3 trillion by 2032, up from $40 billion in 2022.
The question in 2024 is whether the generative AI marketplace will need to license the IP of media conglomerates as much as the latter’s declining models may need those licensing revenues?
Key Takeaway
There is no evidence yet that a generative AI model needs Hollywood’s IP to drive inelastic demand. Even if the market circumstances are otherwise similar to the 1980s video marketplace, they do not shape inelastic consumer demand in the same way.
Total words: 1,300
Total time reading: 5 minutes
Licensing & Rights
Concededly, it is an apples-to-oranges comparison of the business models of generative AI to videocassettes and video rentals in the 1980s. The biggest difference is that a $100 videocassette was effectively a widget, or a gadget with utility, with a use case limited to VCRs. Also, because videocassettes were “widgets”, their spools of magnetic tape were not built for the repeat usage from video rentals. they broke often and needed to be replaced. That also helped boost videocassette sales to video rental stores.
The generative AI model “de-widgetizes” IP —Fable’s Showrunner creates new stories based on each new text-to-video prompt from the consumer—and its use case is not limited to any device. No output is exactly alike.
This caveat is important: Without a “widget” and the constraints around producing it, hard questions emerge about the economics on the supply side and the consumer side. In turn, it is harder to imagine whether sufficient royalties will be produced to keep studios, producers, cast and crew enthused about this new medium. This is true despite AI’s deeper disruptive potential.
In the 1980s, studios leveraged a mix of the scarcity of videotapes and scarcity of ownership of VCRs to charge exorbitant prices for purchasing rentals. Movie rights holders were compensated from each sale of a videocassette and not for each rental. Although the nominal cost was $100, studios and video rental stores negotiated the cost per cassette down to $65 to help drive more sales.
A quote from actor and producer Tom Hanks in the book “Powerhouse: The Untold Story of Hollywood's Creative Artists Agency” sums up why video rentals became so lucrative:
“Every movie, every mom-and-pop chain rental store had to buy at least one copy—two or three if they wanted to make sure they could always have something for their customers. So that meant two or three copies times how many video stores were there across the country, half a million? Suddenly the VHS business was bringing in big money.”
Retail Models
These economics posed other challenges for video rental stores. As Payne writes:
“...several years of experience had taught video store owners that customers believed a fair price to rent one was about $3. This meant stores had to rent a movie 22 times just to recover its cost and many more times to pay expenses and generate a profit.
A New York Times article from 1985 highlighted how stores priced rentals on a daily basis from as little as $1.99 per day ($5.82 in 2024) to $5.95 ($17.39) per day. Lower prices were offered with annual “club memberships” that ranged from $29.99 ($86.87 in 2024) to $75 ($219.25). Late fees ensured that supply met demand (and when Blockbuster emerged at the end of the decade, it pursued late fees more aggressively than any other chain).
But, Payne writes that the need to generate so many rentals during the high-demand period immediately after release “made it impossible for video stores to stay in stock on all titles all the time.” In short, rentals were still expensive for consumers and made the economics of running a video store unattractive.
Two Questions
Two questions emerge from this history:
Without similar pricing power on the supply side, why would media conglomerates expand into licensing their IP to generative AI? And,
Without the economics of inelastic consumer demand, why should a generative AI platform attempt to build a business with licensed IP?
The challenge to answering both questions is that, for all the promise of text-to-video storytelling, generative AI models create an abundance of permutations and not scarcity.
Like early video rental outlets—originally, videocassettes could be rented at electronics dealers, music stores, and other specialty retailers—both multiple entrepreneur-led and tech conglomerate-led platforms are in a position to license IP from movie studios. Some models—like Showrunner and OpenAI’s Sora—are better funded and more advanced than others.
Also, any solution would be unlike a “widget” model. A generative AI platform must scale not only to reduce its operating costs but also to deliver near-real-time iterative improvement. Like YouTube’s or Netflix’s algorithms, a generative AI platform must learn through countless iterations to improve. High-priced studio IPs may not scale often enough to justify the cost of licensing, which is the lesson Netflix learned with Hollywood’s “premium content” model.
There are faster, cheaper and better ways to capturing a competitive share of a $1.3 trillion marketplace.
Inelastic Consumer Demand?
That leaves the question of whether inelastic consumer demand may justify Hollywood studios trying to license their libraries. Generative AI platforms are pursuing Netflix/Spotify-like, consumer-facing software-as-a-service models with low monthly fees to ensure scale of use and recurring monthly revenues. But, as Netflix and Spotify have proven, those fees create a broad base of elastic demand. Any algorithmic platform needs to offer more than just their core value proposition—Spotify recently pivoted to Audiobooks and Netflix to games—to create inelastic demand.
So, there is no evidence yet that a generative AI model needs Hollywood’s IP to drive inelastic demand. Showrunner seems positioned to change that (and more on that next week). Even if the market circumstances are otherwise similar to the 1980s video marketplace, they do not shape inelastic consumer demand in the same way. But, they are similar enough that studios will inevitably need to find a new source of licensing revenues.
It will not be as lucrative.