Rockstar Games gave this title to Netflix for a six-hour exclusive before it was also released on YouTube, where that version has aggregated 26 million views over the last 11 days. It is worth noting that Netflix defines a view as the total hours viewed for a title divided by its runtime, meaning one household watching a program twice would be counted only once. Whereas On YouTube, a view is counted as soon as any video starts to play.

This was a hit for Netflix and Rockstar Games. It marks a significant leap in Netflix’s gaming ambitions, which were launched five years ago when it hired former EA executive Mike Verdu to build the foundation of its games strategy.

My favorite detail of that pivot is that it was announced soon after Netflix’s $200 million bet on comic book IP—a TV series called “Jupiter’s Legacy” from the universe of comic book publisher Millarworld, which it acquired in 2017 for an undisclosed amount—was not renewed for a second season. “Jupiter’s Legacy” was Netflix’s bet that it could compete head-to-head with Disney’s Marvel and Warner Bros.’ DC Comics in comic-book blockbusters. Its cancellation was announced less than a month after its release date.

Effectively, that now-forgotten show and acquisition marked the end of Netflix’s ambitions of becoming the next Disney. Instead, it sent Netflix on a path focused on becoming an interactive mobile and now TV cloud gaming destination. I have argued Netflix has underperformed if not failed in its pursuit of its objective. Others have argued that Netflix has simply failed.

The success of “Grand Theft Auto VI: An Extended Look” suggests Netflix’s audience overlaps with gaming culture more than both I and others have assumed. This presents two questions:

1. Is it the foundation of a broader gaming strategy suggested by Unity’s announcement in late July of “dedicated Engine support to Netflix Games, making it easier for developers to build and deliver games across Netflix’s multi-platform ecosystem on mobile and TV”?

2. After EA went private to compete, is Netflix’s listing as a public company a limitation on these ambitions?

The first question is whether Netflix is building a product people want. The second is whether its size and its shareholders will let it pursue these ambitions.

Past essays related to today’s analysis:

Read more

Runway first generated headlines two years ago for its partnership with Lionsgate, promising to build a new, custom AI model that the studio can use in the editing and production process in exchange for access to Lionsgate’s content library. That partnership hit a wall, as reported by “The Wrap” last September: A model requires “as much data as possible for it to understand context and then to render the right frames, human musculature, physics, lighting and other elements of any given shot”. Lionsgate did not have enough and copyright concerns created a long list of other issues.

Instead of responding by rethinking and streamlining the production process with AI tools, Runway pivoted. Solaris is a screen interface that draws software as you use it. The examples in the Solaris announcement—retail (clothing), restaurant (building a salad), and science (an interactive demonstration of combustion)—are not traditional media.

This contrasts with the emergence of AI tools in pre-production, editing and VFX. If everyone actually wants a new layer rather than better tools, those strategies are aimed at the wrong problem—and the company building the layer is not in the film business at all.

In April 2025, I outlined Fable Studios founder Edward Saatchi’s vision for platforms like Runway and Showrunner in Hollywood. Runway’s tools could be used to “pre-visualize” an entire film. Production executives would watch this pre-visualization, and pre-production would become an interactive process. We are seeing studios fold AI tools into pre-production, editing and VFX.

Runway’s pivot suggests that the thing these true AI visionaries foresee in entertainment requires a new layer rather than better tools.

Past essays related to today’s analysis:

Read more

The fun lens to look at these two announcements is through Netflix’s long-stated objective of 10% of U.S TV consumption (according to Nielsen’s The Gauge). YouTube surpassed the target over two-and-a-half years ago and has soared 40% past it since. Over that same period of time, Netflix has stayed flat and since struggled to reach it.

The yet-to-be-met goal tells two stories. There is a YouTube-centric story of its success in connected TV since Nielsen first started sharing the metric in November 2023. That is largely a story of YouTube’s scale and creator economy model reshaping the smart TV and broader TV ecosystem—and generating over $40 billion in advertising revenue and around $20 billion in subscription revenue (from YouTube Music, YouTube Premium, YouTube TV and NFL Sunday Ticket) annually.

There is also a Netflix-centric story about the broader marketplace which has evolved away from subscriber growth and user engagement, and towards advertising growth and ~$11 billion of net income on $45.2 billion in annual revenue. Both share a key detail: What separates these platforms is not how much premium content they make, but what remains of their businesses without it.

Past essays related to today’s analysis:

Read more

For those who may not be familiar, an MOU is a framework that guides both parties—it is not a binding legal agreement. That is a crucial detail—the best deal that the most powerful rightsholders in entertainment could secure from a Chinese technology conglomerate is a concession that copyright does not apply to AI training and generation.

It also means that everyone has agreed to rules of conduct but not a contract. MPA Chairman and CEO Charles Rivkin labeled these rules “meaningful guardrails on Seedance and Seedream”, and John Rogovin, general counsel of ByteDance, referred to them as “an important framework for continued collaboration as the technology evolves, across a variety of products and platforms.”

As I highlighted back in April, a key problem of studio copyright infringement lawsuits against AI companies overseas is that “U.S. copyright law does not reach conduct that occurred entirely abroad.” Videos generated on servers outside the U.S. are simply not covered by U.S. copyright law—and the 52 infringing videos at issue in the Disney v. Minimax case were created by the studios’ lawyers and not by any U.S.-based creators. This deal confirms that reality—IP protection in the generative AI space is a negotiated set of rules of engagement between studios and popular Chinese AI platforms, but not a legal guarantee.

That problem points to the real question underlying this MOU: if the law could not force ByteDance’s hand, why did it sign?

Past essays related to today’s analysis:

Read more

On July 22, Lavazza released a digital short directed by Francesca Scorsese which follows her father—famous director Martin Scorsese—through New York while he drinks espresso. It is one example of brands importing the operating skillset of a studio and running it inside a marketing budget.

The Hollywood Reporter’s recent cover story, ”The Pillsbury Doughboy Is Coming for Hollywood” details a wave of productions like it—a Prada short directed by Barry Jenkins, a scripted series from Albertsons and Procter & Gamble—and frames brands as “keeping an ailing industry afloat.” But it also presents brands as outsiders stepping into a market with declining opportunities for talent and building studios to create content with that same talent.

This complicates the argument made recently by activist hedge fund Anson Funds to Lionsgate management that third parties will license or even buy libraries of existing IP from studios like Lionsgate. I have previously argued that advertisers are the optimal licensing partners for this IP, especially in an era of generative AI: “Advertisers can license ‘five or six IPs’ from the major studios while figuring out the ‘big swings’ through many ‘little swings’ in the background.”

The story suggests, instead, that brands now prefer to build their own IP and implies some additional challenges for brands as third party customers for licensing.

It also suggests the demand for libraries of IP from brands is not as strong as my past essays argued. This is because brands who might license IP from Lionsgate and other studios are choosing to build new IP by funding original content instead. In April, I predicted brands would “build their own entertainment infrastructure” and shift dollars away from “the traditional advertising value chain altogether.”

Last month I flagged two risks in a world where brands license IP from studios for generative AI campaigns. First, advertisers will only license if IP can still be kept excludable—meaning the owner can prevent creators and advertisers who do not pay for access from using the IP—and be enforced outside the walls of the IP holder’s cloud ecosystem. For example, an advertiser licensing “Teenage Mutant Ninja Turtles” from Paramount Skydance will want Oracle to watermark every output so misuse of authorized content can be traced though not enforced.

It is worth noting watermarking would do nothing to stop a separate, unauthorized model trained on the same characters and hosted on other platforms. If exclusivity cannot be enforced, the thing a studio is actually selling—premium IP with the right to exclude everyone else—becomes unenforceable by default. That presents a bigger, more existential question of whether IP ownership still functions as a business model once any well-resourced actor can train a separate model on the same characters. That is a key reason why studios reacted so strongly to Seedance 2.0’s release in February.

Otherwise, brands will be forced to build their own IP. The precedent of Gulf+Western owning Paramount Pictures suggests they could buy a studio, too.

A second risk is that audiences must care about the IP enough to engage and pursue a direct relationship with authentic, high-quality representations of the IP from a brand. WME head of entertainment marketing Liz Walaszczyk told The Hollywood Reporter that brands now “have more of an opportunity now to be more targeted in terms of the audiences that they’re trying to reach”. Kraft-Heinz CMO Todd Kaplan added that the guiding “philosophy” is how to “create organic conversations that are really opt-in from consumers and things that people want to share with their friends.”

Kraft-Heinz is currently betting on Disney IP to revive sales of iconic brands like HEINZ, Philadelphia and Kraft Mac & Cheese. If third-party IP cannot accomplish these goals, there will be more reasons to build original IP than to license it.

A Marketplace Before Generative AI

Today, brands are opting to build. The question is when, if ever, license or buy enters into that calculus.

I argued the Kraft-Heinz deal reflects two different futures for how generative AI can enter the fold. If Kraft-Heinz “has generative AI and branding savvy” and Disney agrees to use cases of its IP similar to its failed deal with OpenAI's Sora, we will see these ads within the next year. If either falls through, it will reveal how risk-averse, and how far behind, these two giant companies are in the AI marketplace

This leaves an open question the Hollywood Reporter article does not answer: Are brands building because they do not want licensed IP, or because the licensing market is not yet liquid enough to serve them? The distinction matters. If it is the latter, Anson’s licensing thesis for Lionsgate is too early and has obstacles to manifest.

The signal to watch is whether a brand licenses a major franchise character for a generative AI advertising campaign within the next two quarters, especially with the holiday season approaching. If that deal does not happen, the answer is the first one: brands do not need licensed IP, regardless of how liquid the market becomes.

It may be simpler for a brand to buy a studio on the cheap, or build its own with the tools of a new medium.

Past essays related to today’s analysis:

The letter posits that Lionsgate could “enjoy a premium as the owner of thousands of films, given the IP licensing that could come from that library.” Management has “struggled to define that to investors.”

Semafor appears to be the only external media party to read the letter—it is not sharing the full letter, only the core logic of the Anson Funds’ arguments. That said, this logic echoes something I argued back in June about Lionsgate’s equity investment in Runway, the video generative AI company:

“Lionsgate and Runway seem to be taking the traditional studio approach to AI. It is a studio that thinks short-form AI content and new IP with a hot AI platform counts as risk-taking. That contrasts with the Bollywood ecosystem where the entire production model is being reimagined and restructured.”

Anson’s logic is similar but narrower: Regardless of what may be happening in Bollywood or elsewhere, emerging models like Sora and Seedance present the traditional production model with a new calculus of build, license or buy.

Anson’s argument is that licensing is the best path ahead to unlocking shareholder value. The firm understands that there are limits to building in the new calculus of build, license or buy. But Anson does not appear to grasp the emerging “first principles” of the calculus of license or buy.

Who can Lionsgate sell to? Who can they license to? And how will it license it? There are limits to what a standalone production company can do in this marketplace—much less than a relative behemoth like Disney or NBCUniversal.

Past essays related to today’s analysis:

Read more

Author’s Note #2: There were a few copy errors in the mailing which have since been fixed.]

In its FY Q3 earnings call, Disney sounded like a company that finally understands the “inherent contradiction within their pivot to streaming”. Back in March, I argued former Disney CEO Robert Iger’s legacy from an aggressive, multibillion dollar pivot into streaming is “a database at the heart of where human creativity once resided.” Direct-to-consumer software models live and die by their databases of consumer data.

Disney Experiences division is a DTC business driven by a database. Its Entertainment division was not—affiliates (broadcast and cable) and theatrical distribution are passive consumption models—until Disney’s pivot to streaming. After that, Disney Entertainment became a messy hybrid of two passive consumption, profitable businesses in secular decline and an interactive, increasingly profitable (12.9% in Q3, up from 6.6% in 2025) growth business in streaming.

For these reasons and more, I argued in March that new CEO Josh D’Amaro must:

“simplify the role of the database in Disney’s flywheel. The new reality is that creativity has become a commodity in Disney’s distribution model. Marginal value from creativity must be extracted elsewhere across the flywheel. It is operationally and technologically more complex and inefficient to pursue that model across multiple databases.”

Disney management addressed this new reality with three announcements:

1. Data unification

2. A move of consumer products from the Experiences to the Entertainment division, and

3. A content-sharing partnership with TikTok

Data Unification

On the call D’Amaro outlined the company’s path to data unification. He acknowledged that Disney’s multiple acquisitions under Iger had created “disparate data sets that sometimes don’t talk to each other.” To fix that, Disney is “unifying” its consumer data “across the entire company so that we can serve our fans better and then drive lifetime value.”

D’Amaro needs this data foundation for a more ambitious bet he shared back in March: turning Disney+ into a “super app” folding streaming together with theme-park tickets, cruise bookings, games and merchandise into one hub. Bloomberg reported in July that the concept remains an early internal idea, with no concrete build underway.

D'Amaro echoed a sales pitch former CEO Bob Chapek had made years earlier when he said: “There are very few companies in the world with the breadth and richness of data that Disney has across parks and streaming and studios and consumer products.” Chapek saw that overlap as the basis for Disney become a DTC company at its core and an opportunity for a “Disney Prime”-type service.

That all said, D’Amaro also acknowledged in an answer to another question that “There’s still work to be done on unifying the tech stacks between the legacy standalone services, as well as on integrating what have historically been disparate data sets.”

Effectively, he was conceding that Disney had done more work to impress investors than to build the platform that could deliver the promises it was making to those same investors. The super app is a preview of why data unification matters beyond customer service. It is the precondition for a product Disney has not built yet but envisions as table stakes to compete with Netflix and Amazon. Even if it is now headed in the right direction, it has a long way to go.

Consumer Products

Disney announced the day before its earnings call that “Disney Consumer Products will shift the majority of its businesses from Experience over to the Entertainment division, sitting within the Studios.” In its letter to shareholders, management said the move “shift will have strategic and operational benefits by bringing the monetization of our IP through consumer products closer to the studios that create that IP.” It added the move “will better reflect the returns our Entertainment segment is generating from the content it produces and make our Entertainment segment more comparable to peer reporting methodologies.”

A memo from Thomas Mazloum—chairman of Disney Experiences—and Alan Bergman—chairman of Disney Entertainment Studios—explained that the objective of the move is to create “cohesion” across the entire Disney ecosystem “from the very beginning” and extend the relevance of its franchises “for generations.”

Consumer products is a profitable business, generating $560 million in operating income on $1.1 billion in revenue in FY Q3. Nothing in Disney’s portfolio has the same 40% margins at the same scale of revenue as cable had. Combining consumer products into Entertainment boosts operating income by 30% from 11% to 14.3%.

It is also worth considering this move in light of the recent deal with Kraft-Heinz, which “unlocks opportunities for integrated marketing campaigns, digital content and brand-led marketing campaigns across Disney’s media platforms.” The implication is the move makes it easier for the content side to execute this part of the deal. The framing of “majority of its businesses” suggests that the division responsible for the Kraft-Heinz products related to the various Parks and Cruises will remain in Experiences and oversee that part of the deal.

It is reasonable to assume this will be a new model going forward.

Content Sharing with TikTok

Disney announced a pilot with TikTok to “bring an expansive collection of thoughtfully curated Disney-centric fan-created content from TikTok to the Disney+ app.” As part of that pilot, “TikTok will offer their creators access to assets related to hundreds of films and series from Disney’s vast library of enduring franchises and beloved IP.” When the pilot launches, “short-form videos from participating creators who opt-in to the program will live on both TikTok and in Verts on Disney+ and feature characters and stories from across Disney’s brands including Pixar, Marvel, Star Wars, FX and more.”

Disney has found an alternative solution to the sudden collapse of its planned Sora pilot alongside its $1 billion investment in OpenAI. This deal was unthinkable at the beginning of 2025, if not heresy to suggest it at Disney 20 years ago.

As D’Amaro explained on the call, the experiment has a clear DTC rationale. First, it helps with awareness and engagement for Disney+: “It’s important for us as The Disney Company to be out there with the fans, making sure that they’re seeing us, they’re engaging with us.” It also helps with retention: offering native TikTok content on Disney+ will create “a more complete experience on Disney+, and it’s a stickier app. People stay around for longer.”

But, “Work To Be Done”

We can see the pieces in place for the new models that are emerging in the post-cable media marketplace. In moving consumer products to Entertainment, Disney has positioned its IP for brands as both advertisers seeking to create content with its IP and licensors seeking to build merchandise with Disney IP. In partnering with TikTok, it is building creator-led entry points into the Disney flywheel.

These are promising opportunities. That said, the key takeaway from the call is that data unification is a necessary condition for these moves to truly succeed. Streaming’s margin gains came from data streaming already owned, and Disney has not shared the extent to which they came from unification with other divisions. The TikTok and consumer products announcements require exactly that kind of cross-division and cross-platform connection to succeed—which D’Amaro says is still incomplete. The TikTok deal will not work if Disney cannot connect a viewer’s TikTok engagement back to their Disney+ profile to know whether it is actually driving retention. Nor will it work if valuable data from that engagement is not actionable across Disney because there is still “work to be done”.

The consumer products move will mean little if the disconnects between databases interfere with the performance of brands’ ad and merchandise campaigns. The slower and/or more problematic data unification is within Disney, the less progress these deals will make for the company despite its obvious business logic and promise.

That suggests that even if the consumer products move makes it easier for Disney to pursue generative AI ad marketing deals with Kraft-Heinz and other advertisers that use its characters, D’Amaro is acknowledging there is still “work to be done” on “unifying the tech stacks” and “integrating disparate data sets”.

He may be intentionally understating the challenge here, given how much of the reorg's logic depends on investors believing the unification story on its face. Disney is well behind where it should be technologically. That was the case that Wells Fargo analyst Steven Cahall made last month when he suggested that Disney should exit the streaming video business. Even with a deal like the one it made with TikTok, Disney is not positioned to compete effectively with high-volume streaming platforms such as Netflix and YouTube. He also questioned whether Disney releases content frequently enough to control subscriber churn and support long-term margins.

Data unification could help to alleviate those problems and create new business problems. But D’Amaro’s explanation suggested there are operational, cultural and technological disconnects across Disney’s data silos. It may be too far behind the speed of the market’s evolution to truly solve this issue.

Cahall’s arguments suggest Disney has insufficient technological resources to compete. I have suggested it also lacks the technological resources to protect its IP and therefore needs a cloud company as a strategic investor.

The real risk D’Amaro revealed is that its data still does not offer a complete picture of which of its ambitious new initiatives will earn their keep.

Past essays related to today’s analysis:

Implicitly, Unity is messaging that Netflix is changing the way it looks at its distribution architecture. The story Netflix has told about its gaming strategy, to date, is a business logic of “Engagement, revenue, profit drives [the] flywheel”. Recently, that story to investors has been both a careful retreat from mobile gaming and careful entry into the TV gaming space. The former suggested “Netflix is retreating from gaming exactly when gaming architecture becomes the model for AI-era media.” The latter has been a quiet success: According to data from AppMagic, a Sensor Tower Company, the Netflix Game Controller app has been downloaded over 22 million times since March 2026.

To date, Netflix has tightly managed developer access to its ecosystem. It began acquiring game studios back in 2021. It launched both first-party games using its own IP—like “Stranger Things”—and third-party games with the sales pitch “no ads, no fees, no in-app purchases”. In recent years, it opened up its ecosystem to popular games like Rockstar Games’ “GTA: San Andreas”—which delivered nearly 60 million downloads between December 2023 and December 2025, according to AppMagic.

The obvious message of Unity’s announcement is that Netflix will be trading its playbook of one-off curated deals for a platform-scale gatekeeper model, much like Xbox or Steam’s certification systems. By comparison, Sony’s Playstation Store has almost 10,000 titles in its digital store, Microsoft’s XBox Store has over 17,000 games in its store and Steam—an online store for PC games—has over 120,000 playable games.

If Netflix is indeed making this trade-off, it is messaging two changes to its business model. First, it is messaging a more dynamic gaming business model—effectively a mix of B2B (access to the Netflix ecosystem) and B2C. Second—and less obviously—it is seeing that consumers value interactive experiences more than the passive viewing experiences of Hollywood movies and TV series.

These are not small changes.

Past essays related to today’s analysis:

Read more

So the decision facing IP holders in traditional media is build, license or buy. If the logic of that dilemma reads familiar, it is because Eisner’s successor Robert Iger and other legacy media executives faced it when confronted with the competitive threat from streaming in the 2010s. Disney first acquired a one-third stake in BAMTech in 2016 and then acquired a controlling stake in 2017—effectively a mix of licensing and building. Other media companies opted to license commoditized third-party software while Netflix and YouTube iterated on their own platforms.

I wrote on Monday that the emergence of generative AI is “rapidly upending traditional processes and business models for content production, distribution and the legal protections for intellectual property. For this reason, it is harder for anyone to identify when to license IP versus when to build.” In doing so, generative AI is disrupting both traditional licensing models and streaming models.

The open question from Eisner’s framing is where buying fits into this calculus. He argued that IP acquisition is “a means to an end” in both traditional entertainment and gaming:

“The key is after you have five or six IPs, you make a great game that nobody’s ever heard of. It’ll be a giant game. But it’s going to take a while.” From the build, license or buy perspective, Eisner is arguing that buy is a survival tactic and build is a strategy.

Past essays related to today’s analysis:

Read more

At The Paley Center last week, former Disney CEO Michael Eisner framed this crossroads as a familiar pattern in Hollywood. Media companies “possess enormously valuable IP”, but also consistently misjudge how or whether to extend it into unfamiliar formats. His talk focused on Hollywood’s long, troubled history with entering the gaming industry: “The problem wasn’t creative, he argued — it was technological illiteracy: gaming ‘was so technologically driven in the beginning, which was not the strong suit of any of us.’”

This problem of technological illiteracy still persists between Hollywood and gaming. A gaming executive told me in 2023 that it is a generational problem: “They aren’t gamers” (something Eisner appears to have indirectly acknowledged).

Eisner did not discuss generative AI, but the new medium is as “technologically driven in the beginning” as gaming was in the 1980s. The emerging technology is rapidly upending traditional processes and business models for content production, distribution and the legal protections for intellectual property. For this reason, it is harder for anyone to identify when to license IP versus when to build.

He did discuss Warner Communications’ failure with Atari—it purchased the video console owner in 1976, and had success until it reacted to competitive pressures from home computers by oversaturing the market with video game titles. That contributed to the video game console market crashing in 1983. Eisner’s point was that because Warner management had no background in gaming, it made product decisions that did not reach sales expectations. Disney had a similar experience with gaming under Eisner’s successor Robert Iger with Disney Interactive Studios, which lost $200 million per year from 2008 to 2012 and had already begun licensing to third-party publishers like Electronic Arts by the time it shut the division down in 2016.

Past essays related to today’s analysis:

Read more

Yesterday’s multi-year strategic alliance between Kraft Heinz Company and The Walt Disney Company—which “spans foodservice, media, events and more”—seems straightforward.

Kraft Heinz will be the exclusive provider of cream cheese, some condiments and macaroni and cheese at Disney’s North American parks and resorts, and on the Disney Cruise Line. It will be able to use Disney’s characters and stories across 10 of the food company’s brands, including Heinz, Philadelphia and Kraft Mac & Cheese.

As for the “more”, this deal also “unlocks opportunities for integrated marketing campaigns, digital content and brand-led marketing campaigns across Disney’s media platforms.” It is not core to the deal, but as phrased, it opens the door to experimentation in the marketing space.

These two elements of the deal mirror two recent themes.

First, the licensing of Disney’s characters and stories echoes the “INVERSION” strategy being touted by People, Inc. Chairman Barry Diller—a bet on products and services “birthed out” of its “enormous reservoir of content”. Diller aims to own the branded products instead of licensing, but the business logic is effectively the same: Brand equity from popular IP with existing fan bases translates into product demand.

Unlike People, Inc., Disney is able to find win-win scenarios with giant CPG companies like Heinz because of its extraordinary theme parks business model ($10 billion in operating income in 2025) and broader fan bases for its IP. By comparison, People, Inc. has much less leverage.

Second, the integrated marketing deal heads in the direction of where I believe generative AI advertising is headed.

In April, I wrote an essay about how Nielsen’s The Gauge is “broken” and how “the traditional media value chain—produce content, attract audiences, sell advertising against those audiences, measure with Nielsen—is collapsing at each node.”

I cited an essay from Secret Level founder Jason Zada. It argued that brands are now learning to develop original IP, put it in front of an audience, learn in real time and iterate, the same way a tech company ships and improves a product, rather than betting everything on one expensive, slow-moving campaign.

Kraft Heinz is partially doing that in this deal, but is renting Disney’s IP and audience to “learn in real time and iterate”. Zada adds, “The fear of the big swing gets a lot smaller when you can take fifty fast swings first, each one more informed than the last.”

I also argued in April that Netflix’s ad inventory is increasingly worth more than its streaming content. The same logic applies here. “Integrated marketing campaigns, digital content and brand-led marketing campaigns across Disney’s media platforms” enable Kraft Heinz to take many “fast swings” within Disney’s ecosystem and with Disney’s customers.

The upside for Disney is that it reaps the benefits from the licensing deal—We do not know the terms of the deal, particularly whether the licensing revenues are tiered based on success parameters. The downside really depends on the generative AI and branding savvy of Kraft Heinz.

I wrote in May that “All roads of the [conversion] funnel lead to a sale. The economics are obvious to all parties, production cost savings from generative AI are real and brands have every incentive to adopt.” So if Kraft Heinz has generative AI and branding savvy, we will see these ads within the next year. If it doesn’t, that will confirm this is a traditional licensing deal from risk-averse executives, and reveal just how far behind these two giant companies are in the AI marketplace.

On its surface, the deal seems quite traditional. But it will be worth keeping an eye on what the two companies do together around marketing.

Disney’s deal with Sora—which collapsed when OpenAI abruptly killed it for reasons unrelated to the deal—signaled its openness to third parties creating videos using its characters. It is opening a similar door for Kraft Heinz to do so. We are not far from AI-generated ads pairing Disney characters with Heinz ketchup.

Past essays related to today’s analysis:

Co-CEO Greg Peters explained to investors on the earnings why that story has evolved:

“[W]hen I started about 20 years ago, we had one number to describe engagement, hours, just flat hours, no weighting, no adjustments and very similar to how we’ve evolved other metrics in the business. Since then, we’ve gone through about a dozen major iterations of our understanding that.

We get more and more sophisticated because we know, ultimately, it’s combined quality, variety and quantity of engagement that translates into satisfaction and value for members.”

Technology companies like Apple have ditched their vanity metrics when unit sales became disconnected from the actual profit engine, so this pattern is not unusual on its own. But subscription revenue is price times subscriber count—there is no abstraction layer between them the way there is for Apple's unit sales.

Peters’ answer is a wordy concession that Netflix has lost the simplicity of its subscriber and advertising engagement story. In the past, growth was told through subscriber growth. Engagement was told through various reports and Top 10 lists. Its decisions to stop releasing subscriber numbers and to reduce the releases of “What We Watched” to once per year concede that it cannot tell either of these stories anymore.

It does not have an alternative story for investors yet—it relies on a financial story about profit—and they have reacted logically: Its stock price is down 8.4% since the earnings call.

Netflix has long set the market standard for the storytelling of growth in streaming. It is worth asking in the flux that Peters describes: What if they not only do not have a story now, but have lost the story of the media and entertainment marketplace altogether?

Past essays related to today’s analysis:

Read more

“The premise is that if management cannot optimize shareholder value across unrelated assets, then they should either discard those assets or find managers who do understand how to connect them in ways that audiences want.”

The “point” of “portfolio reconstruction” is “not only that the corporate structure of cable bundles no longer matters. It is that the entire corporate entity needs to be rethought.”

Comcast is now pursuing its version of portfolio reconstruction. First, on January 2nd it spun off some cable networks into Versant. The recently announced spin-off of NBCUniversal was its second move. It was the first explicit admission by a large legacy media company that bundling scale-driven and IP-driven businesses under one roof cannot create value in the AI era.

There will soon be three different business models solving three problems related in origin but increasingly unrelated in practice. The paths forward for Versant and Comcast are straightforward: the former can ride out the slow but profitable decline of cable and the latter is a broadband, wireless and entertainment platforms business serving more than 65 million homes and businesses.

Two weeks ago I argued that the best path forward for NBCUniversal is its own version of “portfolio reconstruction”: It should pursue a partnership with a cloud company. This would enable NBCUniversal to license its IP to generative AI users via subscription or per use fees, and also to brands and merchandisers that can quickly produce new long-tail ideas from its library of IP. It also would emulate the Ellison family’s model for Paramount Skydance and Oracle Cloud Infrastructure, but at a smaller scale and without the family control.

A post on X yesterday from Microsoft CEO Satya Nadella directly challenged this recommendation: “In the AI age, the buyer risks giving away knowledge, just in order to use what they bought.”

He adds: “You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!”

Past essays related to today’s analysis:

Read more

As longtime subscribers may recall, my rule is try not to publish when on holiday unless something newsworthy happens. The news of the Comcast split hits on a few recent themes so I am sending an abridged version for my paid subscribers below.]

On Monday Comcast announced that it plans to separate NBCUniversal and Sky from its connectivity (Cable, Broadband, Mobile) businesses.

Comcast Co-CEO Mike Cavanagh told investors that there was no longer a strategic rationale for its “flywheel” model across Cable, Theatrical and Theme Parks: “where previously we believed that the scale and diversification benefits warranted operating these businesses as one company, we now have simply changed our mind about that.”

In a separate memo to employees, he outlined an aggressive vision to “play offense and move with urgency”: “With a strong balance sheet and a public company stock, we will have the resources and flexibility to pursue the growth opportunities we believe in most.”

The not-so-subtle is that a Comcast as large media company no longer can extract marginal value from a library of IP. So, NBCU will pursue traditional paths of mergers and acquisitions to grow the business. But in this market where the value of IP is being aggressively re-imagined by Bollywood, Paramount Skydance and 1.2 billion people using the internet for the first time, what should Comcast do?

Past essays related to today’s analysis:

Read more

A24 partner and A24 Labs leader Scott Belsky (the former Chief Product Officer of Adobe) told The Wall Street Journal that the objective is “better uses [of AI] that preserve creative control and support risk-taking.” He added that the new tools “won’t look anything like the prompted generation type of AI that people feel uncomfortable with.”

But the most notable detail for our purposes is this sentence: “Their multiyear, nonexclusive deal doesn’t give Google access to A24’s data, including its film and television library.” Instead, it is a bet that the studio’s roster of artists and auteurs will gravitate towards the tools.

A24 is known more for “arthouse gems” than blockbuster IP—so this detail makes sense. Also, A24’s recent success with YouTube creator Kane Parsons’ horror movie “Backrooms”—which just had the biggest opening in A24 history at $81.4 million on a $10 million budget—is also on the vanguard of “Internet Collaborative IP“. It understands that a next generation of storytelling is emerging from creators working with their audiences to build and improve stories.

The article reads like Google making a conservative bet on something Runway CEO Cristóbal Valenzuela recently described as a “Cambrian explosion of creativity, weirdness, new media, and new forms of expression and art” that is “just around the corner.” However, when viewed through the lens of “Internet Collaborative IP”, the deal establishes Google in this emerging marketplace for both generative AI and Hollywood: Google’s YouTube is a core engine of ideation, community and distribution of “Internet Collaborative IP” before a studio acquires the IP; and DeepMind now has tools within the studio’s pre-production side and the production processes.

With A24, Google is building the early stages of a seamless path from YouTube to theatrical distribution for content created by YouTubers like Parsons, Curry Barker (“Obsession”) and Markiplier (“Iron Lung”) and generative AI creators. That pipeline effectively looks like:

Creator → YouTube → Community → IP → Studio → Production Tools → Theatrical → Paid Downloads (YouTube) and/or Subscription Streaming (Netflix)

In this model, Google has a role in everything but the last two stages—and the real “meat” lies in the first three stages.1 It is worth noting it enables paid downloads of YouTubers’ movies in post-theatrical windows instead of subscription streaming.

Looking at this chain, we can immediately see the problem for both traditional streamers and Netflix: Together, Google and A24 will capture more value across the chain. Netflix and other streamers who buy A24 movies and TV shows will be left waiting for the “scraps” of streaming distribution.

Past essays related to today’s analysis:

Read more

Last week’s essay on Lionsgate’s deal with Runway concluded:

What does it cost risk-averse executives to truly take on risk? Bollywood’s answer is: far more than they are imagining now.

I had a few conversations and exchanges over the past week that left me thinking my focus on Hollywood is understating the magnitude of this challenge. That if we look at the future of AI in Hollywood from the outside in—like what is happening in Bollywood or through the lens of the Disney v Minimax lawsuit—the industry will never move fast enough to catch up.

The business culture was built to create and support TV and movies for an incredible once-in-a-lifetime profitable business model (cable at 40% EBITDA margins and broadcast at 20%) and a hit-driven model (theatrical at 3% to 6% operating margins) built for passive viewing. Streaming emerged as a medium for interactive viewing. The most successful and profitable once-in-a-lifetime profitable business model has emerged on interactive platforms Netflix (67% EBITDA margins)—but not on YouTube (estimated to have “mediocre” margins), while their legacy media company streaming competition are either losing money or barely breaking even (Disney at 5% operating margins for streaming).

The obvious question for any corporation is where to find additional profits as the cable industry declines and theatrical struggles to recapture COVID-era attendance.

Generative AI enters that equation in two ways: Replacing existing cost-centers like traditional visual effects (VFX) or replacing live productions altogether. The former has been happening in Hollywood since 2023, where ~75% of entertainment industry executives were already using AI to remove, reduce, or consolidate jobs. More recently, Netflix bought Ben Affleck’s AI studio InterPositive “to accelerate the use of artificial intelligence in its filmmaking”.

Netflix’s last engagement report is the missing piece that explains the significance of these moves. Its What We Watched in H2 2025 showed total hours viewed in the second half of 2025 increased by only 2 percent. Research firm MoffettNathanson estimates that Netflix’s global user base grew by 10 percent last year, “which translates to an 8 percent decline in daily engagement per subscriber.“

Both that acquisition and the Lionsgate-Runway partnership prove that Hollywood can buy AI capabilities. However, the recent guild negotiations have proven that its core participants will never be unethical or shameless enough to compete with AI studios and corporations from countries like India, China who do not protect copyrighted materials in AI or lean towards embracing AI technology over traditional forms of human creativity. In Singapore, a 2021 Copyright Act explicitly allows commercial AI training on copyrighted content in a way that U.S. and EU law does not.

The owners of studios want and need to find more profitable business models, especially in online media. Rapid subscriber growth normally delivers less engaged consumers—but an 8-point gap between hours growth and subscriber growth suggests something more structural for Netflix—paying subscribers are finding alternative sources of entertainment. A drop in engagement is the opposite of what a platform increasingly relying on ads needs to convey to the marketplace.

As I wrote in March, Netflix now has plenty of competition from both emerging generative AI apps like ChatGPT and Claude, and especially vertical video streaming:

“The challenge for Netflix is not only that this new format is competitive with the 27-year-old incumbent in such a short period of time, but that its weekly and monthly spend suggests consumers value the new format more.”

This is a signal that there is a substitution underway and being accelerated by the 1.2 billion new users who gain access to the internet between 2022 and the end of 2026. Those two signals point to the deeper challenge: Hollywood is moving too slowly, and not relative to Bollywood’s or generative AI’s production costs.

Instead, it is racing against the rapidly evolving attention habits of an audience that has not yet decided what “good enough” means—and is making that decision with less and less Hollywood content in the room. Buying AI capability inside a guild-constrained culture for long-form content seems to be an insufficient solution for rapidly changing audience preferences around storytelling.

From Passive Viewing to Internet Collaborative IP Storytelling Models

What to do instead?

Obvious signs point to vertical video and generative AI as alternative content formats to be embraced as the 1.2 billion increasingly consume storytelling within them. Hollywood studios appear more eager to embrace the former than the latter.

“Internet collaborative IP”—which I wrote about earlier this month—appears to be a more viable solution. It is “IP that is for the people by the people… an interactive story that was built over time that you feel a part of if you were there which makes you want to go and see it and celebrate it because the people made the story.”

The approach marries both the interactivity of online storytelling consumption and the passive consumption of theatrical. It suggests the answer is not for studios to produce long-form content faster or cheaper with AI across televisions and movie theater screens. Instead, it is to rapidly embrace content formats the 1.2 billion are already choosing and to do it before their habits calcify into something Hollywood had no hand in shaping.

To date, we have only seen three movies from YouTubers— “Iron Lung”, “Backrooms” and “Obsession”—emerge and succeed at the box office. That seems likely to grow before any of the guilds shift their embrace of AI. That is because for studios, the corporations that own them and the guilds that have contracts with them, the production budgets can be extraordinarily low (budgets for “Obsession” and “Iron Lung” were both under $1 million) and human creativity can be both protected and thriving.

AS I previously argued in both “Make Ads or Make Stories” and “Bottlenecks and The Greenlight Gap, advertising is the only path where the economics for both creators and distributors make sense in a marketplace where generative AI scales. There is a second solution—one that involves brands, advertising and content partnerships—which I will take up in a future essay.

Past essays related to today’s analysis:

There is an obvious streaming angle to this—Fox has already had success with the free, ad-supported platform Tubi. It surpassed $1 billion in revenue for the first time in FY 2025 and now comprises 23% of Fox’s total advertising business. According to Nielsen’s most recent The Gauge, it is the second most-watched free-streaming service in the U.S., behind only The Roku Channel. Acquiring Roku is doubling-down on that business and boosting Fox’s share of U.S. TV consumption by platform from 2.2% to 5.2%, which would equal Disney’s current share across three apps and compete for third place behind Netflix.

There is also an obvious U.S. advertising angle to this: Combining Roku with Fox’s broadcast and cable networks—and Tubi—would give them the third-largest share of U.S. television viewership by distributor(10.2% as of March 2026), behind only Disney (10.6%) and YouTube (13.2%). Fox will be able to offer advertisers competitive scale and sophisticated ad-targeting.

There is a less obvious generative AI marketplace angle to this. For the past two years I have been researching and analyzing how and why AI restructures who captures value in entertainment. My answer has been: whoever owns both the content and the infrastructure (The Ellisons controlling Paramount Skydance and Oracle, Amazon owning MGM Studios). Fox will own two key points of infrastructure:

1. 100 million connected TV households worldwide using a Roku streaming player, a first-party Roku-made TV, or “one of the many Roku TV models built by” its global hardware partners

2. Roku’s advertising platform and Operating System which enables them to capture revenues from digital advertising and distribution of third-party streaming services

But what business will a combined Fox and Roku be in when the streaming model is hitting headwinds, and generative AI creators are more incentivized to produce content for consumers via advertising than original content?

Past essays related to today’s analysis:

Read more

Recent conversations with people in Hollywood point to a shift in mentality towards AI from both studios and guilds. Some highlighted the upcoming IPOs for OpenAI and Anthropic as making AI seem less abstract and more an inevitable reality. Some pointed to few, if any, signs of vitriol between studios and guilds in the current negotiations. All suggested that various mindsets of fear and resistance I heard in conversations as recently as six months ago were disappearing.

Similar signals of a thaw, if not a vibe shift, have emerged in the past few days. Lionsgate announced yesterday it is taking an equity stake in Runway, the generative AI video company. It expands on a partnership launched in 2024. The deal includes using Lionsgate’ existing catalogue of intellectual property for an AI short-form series and starting a new development program to create new intellectual property using AI.

I previously wrote about Runway’s partnership with Lionsgate last year and then again in February about the range of “unforeseen complications” that emerged: The “limited capabilities that come from using just Runway’s AI model” and “copyright concerns over Lionsgate’s own library and the potential ancillary rights of actors”.

This deal seems to solve for both: Short-form series appears to be a business opportunity that avoids the challenges encountered incorporating Runway into pre-production. New IP obviates the thorny questions around copyright and compensating rightsholders of old IP (in theory).

Risk-Taking

This deal is emerging both against the contrasting backdrops of the guild negotiations and Bollywood’s aggressive embrace of AI. Through the lens of the former, it reads like aggressive and possibly transgressive risk-taking —the possible use of existing franchises and characters will create inevitable exposure to public-backlash around labor, likeness, authorship and intellectual property. It will also be subject to the final version of SAG-AFTRA’s ​​12 basic ground rules for the use of digital replicas in its tentative agreement with AMPTP.

Through the lens of the constraint-free “Wild West” of Bollywood, the deal seems like a half-measure. A smaller, traditional studio is embracing AI for short-form content and new IP, only. The partnership is built to aggressively use AI to push new movies into theaters, revise old ones or create movies based on public domain IP like mythology or The Bible.

Disney Precedent

In some ways the Lionsgate-Runway deal echoes Disney’s abruptly-cancelled deal with OpenAI. Last December, Disney and OpenAI announced a surprise three-year licensing agreement: Disney would license 200+ characters to OpenAI’s Sora platform, invest $1 billion and become a “major customer” of OpenAI’s APIs “to build new products, tools, and experiences” Depending on who you asked, that deal was perceived as Disney taking a significant risk, or as I argued, securing the optics of a significant risk but really investing in operational cost efficiencies. Now that the deal is dead, Disney never got to answer that question.

Runway recently raised $315 million in Series E funding at a $5.3 billion valuation to “pre-train the next generation of world models and bring them to new products and industries.” Lionsgate is sending a similar message to shareholders that Disney did: Whatever the perceived costs or risks of AI may be to the creative industry, shareholders will see inevitable financial upside from a legacy media company hitching its wagon to a fast-growing AI platform.

But Is This Enough Risk?

The Eros recut of a 2013 hit movie “Raanjhanaa” proved that Both theatergoers and 958 million active internet users in India are increasingly forming their entertainment expectations around AI-generated content. Hollywood and Bollywood compete for the same customers on the same digital platforms. A creator at the Artist & The Machine Summit framed it best: Asia and Africa are leveraging AI to change the scale and speed at which content is being produced and monetized. With hundreds of millions more arriving online across Africa and other territories getting internet access for the first time, this means a conservative solution that works for the Hollywood ecosystem will struggle to compete globally.

More risk must be assumed by a Lionsgate and Runway partnership to produce content faster and cheaper in order to compete in a global marketplace. Monday’s essay suggested that the model also needs to be “internet collaborative IP”, which is “IP that is for the people by the people essentially. It is an interactive story that was built over time that you feel a part of if you were there which makes you want to go and see it and celebrate it because the people made the story.”

That model is the antithesis of a studio model. It is more like a creator model. It captures the human creative collective that has built Hollywood. It is also its antithesis—audience input eliminates the need for a team of “expert” creatives driving the creative. That makes the model cheaper.

Instead, Lionsgate and Runway seem to be taking the traditional studio approach to AI. It is a studio that thinks short-form AI content and new IP with a hot AI platform counts as risk-taking. That contrasts with the Bollywood ecosystem where the entire production model is being reimagined and restructured.

This raises a difficult question: What does it cost risk-averse executives to truly take on risk? Bollywood’s answer is: far more than they are imagining now.

Past essays related to today’s analysis:

Collectively, both are grossing more than Disney’s “Star Wars: The Mandalorian and Grogu”, which has earned $145 million worldwide on its opening weekend and $294 million since May 22—and on a budget of $165 million.

There are obvious explanations for the financial success of the films: production studios like Focus Features—which distributed “Obsession”—Blumhouse—an executive producer on “Obsession”—and A24—which produced “Backrooms”—are revolutionizing the horror genre both creatively and financially. Horror’s popularity as a genre partially explains these box office successes.

The deeper story here is that YouTubers have individually and collectively outperformed a major studio’s IP juggernaut. The broad brushstrokes of that story map to my argument that the new media marketplace is “more McLuhan than Christensen”: the invisible “groundrules [sic], pervasive structures and overall patterns” of the internet empower creators to have a more interactive relationship with their consumers.

But theater screens are not interactive like a touchscreen or keyboard—they are physical spaces outside of and not contained by Google’s vertical ecosystem. There is only one previous instance of a YouTuber finding success in theaters: In January, Markiplier released his adaptation of the video game “Iron Lung” and grossed $50 million worldwide on a budget of $3 million.

A clearer answer emerged over the weekend from YouTube creators Colin & Samir. They published a video that pinned the success of “Obsession” and “Backrooms” on what they called “internet collaborative IP”:

“This is a story that is built by this generation. It is not IP like Star Wars or Jurassic Park or Spider-Man that is already pre-existing and that the studios are telling us we should go and see. It is IP that is for the people by the people essentially. It is an interactive story that was built over time that you feel a part of if you were there which makes you want to go and see it and celebrate it because the people made the story.”

Their broader point is that “ for all of the conversation about how Hollywood is dying, sure Hollywood is not the same that it was, but it’s [not] dying. [I]t’s just going through an evolution”. An emerging calculus of community interactivity and “great storytelling” is key to the evolution, and that was also a key theme to last week’s stories.

Past essays related to today’s analysis:

Read more

Earlier this week I had dinner with a subscriber who suggested I look at Amazon’s announcement last week at “AI on the Lot” through the lens of the pre-Paramount decrees studio system.

In the 1930s and 1940s, the “Big Five” Hollywood studios were vertically integrated across production, distribution and exhibition. An antitrust ruling from the Supreme Court in 1948’s United States v. Paramount Pictures, Inc. forced the studios to sell off their theater chains and also outlawed additional anticompetitive tactics with independent theaters. In 2020, U.S. District Court Judge Analisa Torres ended the decrees in response to the U.S. government’s request to do so. She noted:

“In today’s landscape, although there may be some geographic areas with only a single one-screen theater, most markets have multiple movie theaters with multiple screens simultaneously showing multiple movies from multiple distributors. There also are many other movie distribution platforms, like television, the internet and DVDs, that did not exist in the 1930s and 40s.”

Studios used to control what audiences saw, when they saw it and where—that is no longer true in today’s diverse entertainment marketplace. Past licensing tactics—like block booking, circuit dealing, and resale price maintenance—would be analyzed in modern antitrust law under the rule of reason: “evaluating the specific market facts to determine whether a practice’s anticompetitive harm outweighs its procompetitive benefits.”

The conventional response to Amazon’s announcement is to quote Clayton Christensen’s disruption theory—new technology enables new vertical integration, and integrated players win early markets. The logic behind Torres’ ruling—that the technological constraints around distribution had transformed the market—is Marshall McLuhan’s “the medium is the message” in legalese. Amazon’s announcement at AI on the Lot, Paramount Skydance’s partnership with Oracle and Google’s Veo 3-to-YouTube-Shorts pipeline all suggest that within only six years, the technological constraints around entertainment distribution had changed exponentially again.

Torres’ analysis already seems outdated.

Vertical Integration in AI

Last December I wrote about the generative AI marketplace (diagram above), where the Ellisons control both content and infrastructure. Amazon enjoys similar vertical integration with MGM Studios and AWS. At AI on the Lot it outlined how its studio, generative AI tools and cloud infrastructure could support animation projects from “people who are leaning into [generative AI] or curious.” There were two other implicit parts to Amazon’s vertical integration pitch: Over 200 million Prime Video subscribers worldwide and “almost 300 million” Fire TV devices “purchased around the world”.

Vertical Integration But No Hardware

Paramount Skydance—via its partnership with Oracle and likely merger with Warner Bros. Discovery—will have similar vertical integration: Paramount and Warner Bros. IP, Oracle’s AI tools, Oracle’s Cloud Infrastructure and distribution via streaming and cable and broadcast networks. However, they will lack the hardware distribution end point.

Vertical Integration But No IP

Google is in a similar position with Gemini (Veo), Google Cloud, YouTube and 300 million active Google TV devices worldwide. Meta has the generative AI tools, proprietary cloud infrastructure, its Family of Apps (Facebook, Instagram, Messenger and WhatsApp) and hardware devices (Meta Quest, Meta Glasses). But both lack the proprietary IP of Paramount Skydance and Amazon’s MGM.

IP and Streaming But No Integration

Netflix owns IP and streaming, but owns no cloud of its own (it runs on AWS) and no hardware (it transferred an internal streaming device project to Roku in 2008 rather than launch it as its own product.). The question is what it will do to remain the dominant global subscription streaming service. It could be the exception that tests the rule, or an example of how full-stack integration is not the only path to market power. There are more emerging market signals it is moving in the direction of generative AI content, suggesting something will have to change.

The New Vertical Integration

The generative AI marketplace is in the early stages of a new wave of disruption and the key players are in the early stages of new versions of vertical integration. Those that can assemble a full stack will occupy the position the pre-Decrees studios held. Because this new vertical integration will exist across new technologies that are different from traditional theatrical distribution, it is not simply a reconstitution of the old one.

The pre-Paramount Decrees studios controlled what audiences saw, when they saw it and where. The new integrated companies have more far-reaching control—the cloud infrastructure that hosts the training data and the content, the AI pipeline that generates new content with studio IP, the algorithm that selects the content for the user and the device that delivers it. The constraints are mostly invisible, embedded in the medium itself.

Because these models are global, the new vertical integration may be more pervasive than theater ownership ever was. It operates through technological constraints that courts are being forced to understand and IP laws which vary depending on region and/or territory. That also implies that antitrust law will struggle to regulate this.

Studios used to own the theaters. If they did not want your movie on a screen, it did not get on a screen. Now the control sits further back. The companies that own the AI models now decide what can be made in the first place.

Past essays related to today’s analysis: