Matthew Belloni on the Ellison media empire: Hollywood's production exodus, AI's limited role, and why the Sopranos won't land on Netflix

Summary

Matthew Belloni maps the structural collapse of traditional Hollywood and explains why the Paramount-Skydance-Warner Bros. merger closes fewer doors than it appears to open for filmmakers pitching original work.

The five-year reckoning started with the end of peak TV around 2020. Netflix's profitability turn forced a content recession that rippled through the industry: fewer productions, fewer jobs, and now a systematic flight of filming to cheaper jurisdictions. The UK offers 40% production rebates and subsidizes above-the-line costs for actors, directors, and writers. Amazon's Project Hail Mary, a $248 million gross-budget film, nets out at under $200 million after incentives. A Netflix deal would have left Warner's theatrical distribution untouched, since Netflix doesn't operate movie theaters. Paramount, by contrast, now owns two film studios and two television studios. Paramount claims it will release 30 theatrical movies annually, an increase, but merger history suggests output contracts downstream and fewer jobs follow.

Netflix and theatrical releases

Ted Sarandos' stated openness to theatrical windows is theater-as-marketing, not a business model shift. Netflix's best marketing platform is Netflix itself. The streamer occasionally puts films in cinemas as promotional devices, but Belloni argues Netflix's model is designed to kill traditional theatrical release windows. Ten years ago, a Matt Damon-Ben Affleck film would have had a robust theatrical run before home video. Not anymore. The Sony Pictures head recently told Belloni that you don't build franchises or stars on Netflix—you build them in theaters with aggressive marketing campaigns that create awareness for sequels years later. Netflix franchises are invisible to the broader world. Amazon, by contrast, is doubling down on theatrical releases this year, betting that creating demand in cinemas first and then moving to Prime Video is the path to durable franchises.

Traditional studios need theaters because the revenue waterfall starts there. A $50–100 million global marketing campaign exists to drive theatrical profitability. Then comes premium video on demand, subscription windows, TV, free TV, and international syndication. Warner Bros. fetches a $111 billion valuation not because of last year's theatrical run but because of a 100-year-old library exploitable across every window forever.

Monetizing the catalog

With $79 billion in debt, cash flow matters immediately. Warner Bros. Discovery faced similar pressure post-2022 merger and has licensed hundreds of titles to Netflix. At one point, six of Netflix's top ten movies were Warner Bros. titles. Sony just renewed its pay-one deal with Netflix for $7 billion over a long period. That's the window Netflix coveted when it pitched buying Warner Bros. Paramount now faces the same pressure to monetize its catalog fast. The Sopranos remains the one marquee title HBO has kept, but debt service may force Paramount's hand to license it to Netflix.

The Ellison factor

Larry Ellison's $111 billion backstop made this deal possible. His daughter Megan Ellison built Annapurna Pictures into a prestigious auteur-film factory in the early 2010s, landing two Oscar nominations for best picture in a single year. But she burned through capital chasing artistic prestige without the 100-year library to absorb losses. Annapurna makes films from scratch; legacy studios have libraries. When Annapurna's losses mounted, Larry shut down the spending spree. Annapurna still exists but operates at a fraction of its former scale. Larry will indulge his children's Hollywood ambitions up to a point. David Geffen may test those limits with this $111 billion bet, though Warner Bros.' proven asset base—the library—is fundamentally different from Annapurna's model of creating a portfolio from zero.

Redbird Capital and PE firms Apollo and Blackstone are providing debt financing. Middle Eastern sovereign funds, Saudi and Qatari, were initially floated but did not appear in the final press release. That gap matters for understanding the deal's backing.

A24's scaling challenge

A24 has built one of the few meaningful studio brands in modern Hollywood. The A24 brand registers more strongly with moviegoers than Warner Bros. itself. A24 succeeded by betting small on auteur filmmakers and cultivating direct-to-consumer brand loyalty through branded merchandise, not just films. But A24 has now raised capital multiple times and hit a $3.5 billion valuation. Success at scale forced it bigger. Marty Supreme cost $70 million; The Smashing Machine cost significantly more than typical A24 fare. The studio still buys small festival films, but the portfolio is now weighted toward bigger budgets. The question is whether A24 can scale without becoming what it succeeded by not being. Many small studios have faltered when trying to compete directly with legacy players.

AI in Hollywood

Hollywood studios are using AI for process optimization and workflow efficiency. For content creation, the toe is barely in the water. Animation divisions are moving most aggressively, with AI automating storyboarding and the physical production of animated images once character design is locked. Studios are treading cautiously everywhere unions have jurisdiction. No studio is greenlighting AI-written scripts. Screenwriters use AI as a tool to research or draft, but the guilds demand human writers be paid for any script greenlit. Actors can be used to enhance or reshoot a scene, but the people involved must be compensated. The next labor negotiation will be the real test, with studios trying to chip away at those protections and guilds trying to lock them down further.

The power of celebrity likeness remains insurmountable. The viral Seedance video of Brad Pitt and Tom Cruise fighting worked because they were Brad Pitt and Tom Cruise. Two anonymous AI-generated fighters are stock footage. That asymmetry explains why Hollywood can still protect actor likenesses from AI replication. A two-tier content ecosystem will likely emerge: AI-generated content at lower perceived quality and lower prestige, and union-backed Hollywood work at higher prestige and higher cost. The Disney-OpenAI deal signals the direction, but it notably excludes voices and likenesses of real people. Even the Luke Skywalker in the deal is animated, not voiced by Mark Hamill. That constraint limits the appeal of AI-generated character work until the technology either breaks that constraint or studios decide the prestige hit is worth it.