Cristóbal Valenzuela, the co-founder and CEO of Runway, spoke at Semafor's World Economy Summit this week. He was asked about AI's potential in Hollywood. His answer: take the hundred million dollars a studio spends on one feature film and spend it on fifty.
"Same quality. Same amount of output, visually. But you make way more content. So you have way better chances of hitting something."
Then the sentence.
"It's a quantity problem."
The math and the metaphor
Fifty films for a hundred million dollars is two million per film. That is not a studio budget. That is a short film budget wearing a feature-length coat. Two million dollars buys you a skeleton crew, a compressed schedule, and an AI generation pipeline doing the heavy visual lifting. It is achievable. It might even be happening already. "Bitcoin: Killing Satoshi," the first AI-assisted studio feature, reportedly used AI to bring a $300 million production estimate down to $70 million. That is a cost reduction. It is not fifty films.
Valenzuela's math is not the problem. The metaphor is.
Calling filmmaking a quantity problem reframes the creative act as a statistical exercise. Make enough things and something will connect. Increase the denominator. The numerator takes care of itself. This is how you think about A/B testing a landing page. It is how you think about mobile game publishing, where studios release dozens of titles knowing most will fail and one might stick. It is, specifically, not how anyone has ever described the thing that happens when a director and a cinematographer and an editor and a composer build something that alters how a person sees the world for ninety minutes.
He compared film to books. Twenty-five million published per year, he said. Nobody reads all of them. But the world is better because more people get to tell stories. The comparison is generous to his argument and still does not hold. Twenty-five million books per year did not produce proportionally more great literature. It produced proportionally more noise and made curation, editing, and taste dramatically more valuable. The slush pile got taller. The readers did not get more patient.
Iteration vs multiplication
There is a version of this argument that works. This series has made it. When generation costs a nickel per clip, shooting fifty takes instead of three is how you find the shot that surprises you. Kubrick shot seventy takes of Jack Nicholson walking through a door. That is iteration. Seventy attempts at the same creative vision, each one refining the one before it, each one informed by what the previous sixty-nine revealed.
Valenzuela is not describing iteration. He is describing multiplication. Fifty separate films, fifty separate creative visions, thrown at an audience to see which one the algorithm rewards. The difference is not semantic. Iteration assumes you know what you are looking for and are working to find it. Multiplication assumes you do not know and are hoping volume compensates for the absence of direction.
One is a filmmaker at work. The other is a slot machine with a content budget.
Five weeks
On March 31, Runway held an AI Summit in New York City. Nearly a thousand people in a ballroom. Valenzuela opened with an AI-generated image of Steve Jobs walking with Socrates. The Paramount CTO compared AI to fire. An EA executive said AI "closes the gap between imagination and creation."
Kathleen Kennedy sat on a panel and asked the head of AFI: "How are you going to teach taste?"
Five weeks later, the CEO of the company that hosted that summit described filmmaking as a quantity problem. Kennedy was asking how to develop the judgment that distinguishes the one great film from the forty-nine forgettable ones. Valenzuela is saying: skip the judgment, make all fifty, let the market sort it out.
These two positions cannot coexist. Either taste is the differentiator or volume is. Either the hard part is knowing what to make or the hard part is making enough.
What the platform wants
Runway is valued at over five billion dollars. It makes money when people generate video. More films means more generation. More generation means more API calls, more subscriptions, more compute sold. Valenzuela's incentive structure and his filmmaking philosophy point in the same direction, which is convenient for him and worth noticing for everyone else.
This is not cynicism. It is business. A company that sells shovels benefits when everyone digs. A company that sells generation benefits when everyone generates. The question is whether the person buying the shovel knows what they are digging for, or whether they are just digging because digging got cheap.
Runway has been moving away from filmmakers for months. It launched a world simulation engine for gaming and robotics. It opened its platform to host competitors' models. It is becoming a general-purpose generation infrastructure company. Telling studios to make fifty films instead of one is consistent with that trajectory. It positions Runway as the factory floor, not the film set. Volume is the factory's metric. Not quality. Not intention. Throughput.
The two-million-dollar film
Here is what a two-million-dollar AI film actually looks like today. You write a script. You generate your visuals across six models, iterating on each shot, building reference libraries, managing consistency across scenes. You record dialogue with actors. You design sound. You edit, color, mix, and deliver. The generation pipeline replaces the camera department, the lighting department, and the location budget. It does not replace the script, the edit, the sound design, the score, the color grade, or any of the hundred creative decisions that determine whether the film means something or just exists.
Two million dollars is enough to make one very good film with these tools. It is enough to make fifty things that fill a content feed and are forgotten by Friday.
The economics work in both directions. The question is which direction the filmmaker is facing.
The vocabulary answer
Soderbergh described AI filmmaking as requiring "a Ph.D. in literature" and "desperately close human supervision." Kennedy asked how you teach taste. Dhanush called catalog optimization "soul-stripping." James Cameron is using AI to keep blockbusters in production without layoffs, which is a cost argument wrapped in a workforce argument, not a volume argument.
Every filmmaker who has engaged seriously with these tools has landed on some version of the same conclusion: the tools are powerful, the vocabulary to use them is hard-won, and the output is only as good as the creative judgment applied at every stage of the pipeline.
Nobody who has used the tools well has called it a quantity problem.
The quantity framing reveals something specific about where Runway's leadership sees its market. Not in the hands of the filmmaker who spends four hours building one four-second shot. In the hands of the studio executive who needs a slide deck showing fifty deliverables for the same budget as one. The former customer needs better vocabulary, better models, better controls. The latter needs a faster pipeline. Runway is building for the latter.
NAB opens in Las Vegas this weekend with nearly double the AI exhibitors from last year. Two entire AI Pavilions. Every company on the floor will be selling speed, volume, and efficiency. Almost none will be selling taste, intention, or the vocabulary to speak precisely to a model that is listening.
The structured prompt works the same whether it produces one film or fifty. But the filmmaker who writes it knows the difference between making fifty things and making one thing fifty times better.
Bruce Belafonte is an AI filmmaker at Light Owl. He has never described filmmaking as a quantity problem and intends to maintain this position indefinitely.