AI Startups Raise Billions to Rewrite How Hollywood Makes Movies

A 15-person team at AI video startup Higgsfield AI produced a 95-minute feature film in under three weeks for $500,000 — and four out of every five of those dollars went to compute, not people.

The movie, Hell Grind, debuted this spring in Cannes, France. It drew attention not for an A-list star or an Oscar-winning director, but because all of its synthetic actors and visuals were entirely AI-generated. The budget breakdown signals where the economics of filmmaking may be heading: compute as the dominant cost line, with personnel reduced to the remainder.

Higgsfield AI's co-founder and chief strategy officer, Mahi De Silva, argues that AI will not eliminate the craft. Future films will still require directors, screenwriters, set and costume designers, and even actors, he says. "What we have found is that filmmaking is still an art," De Silva says.

The startups pitching investors tell a more aggressive story. They envision AI disrupting almost all of the creative work that goes into filmmaking, from Hollywood to Bollywood — a shift that would almost certainly cost jobs for production crew and visual-effects workers.

The money is already moving

Higgsfield AI, founded in 2023, is having a breakout year with investors. In August it raised $400 million in a Series B at a reported valuation of $5.4 billion — a fourfold jump in eight months. The startup had raised $80 million in January at a $1.3 billion valuation.

Runway raised $315 million in a Series E round this year that valued the startup at $5.3 billion, and it has sold an equity stake to Lionsgate, the studio behind The Hunger Games and John Wick, to create content with AI.

Luma most recently raised $900 million in November 2025, with backers including the Saudi AI company Humain and chipmakers Nvidia and AMD.

McKinsey estimates that by 2030, AI could influence up to 20% of original content spending on films and TV. The consulting giant says the shift opens opportunities for small studios, more personalized projects and short-form content formats.

Labor has already drawn the line

AI's role in filmmaking became a main provision in a new four-year contract ratified between major studios and SAG-AFTRA, the labor union representing 160,000 actors, broadcast journalists, news writers, editors and other artists. The agreement put strict guardrails around how AI could be used and when content and payment would be required.

Audiences may also set limits of their own. In 2026, some of the biggest box office hits, including The Odyssey and Obsession, were made using practical effects. Moviegoers were less keen to open their wallets for films that heavily relied on computer-generated effects.

What the technology can actually do

Higgsfield AI's technology centers on Cinema Studio, an AI-powered filmmaking platform that gives creatives access to 33 AI models, including OpenAI's ImageGPT, Google's Nano Banana and Seedream from TikTok parent ByteDance. "We're constantly evaluating new ones," De Silva says.

The company is also calibrating how much AI to use. For this year's Cully Hill Boys, Higgsfield AI legally licensed the likeness of four actual humans, including UFC fighter Israel Adesanya, for the AI-created characters.

Runway's capabilities have advanced quickly since it debuted its proprietary video generation model, Gen-1, in early 2023. Early use cases were limited to restyling existing video. The company's Gen 4.5 model, which debuted in December, can generate videos from written prompts. "There really isn't a part of the production lifecycle that can't benefit from AI," says Jamie Umpherson, Runway's chief creative officer.

Runway's recent projects include work with animator Jeremy Higgins on the animated series Mars and Siv, and a short film, Rules of Horse…for Dummies, produced in under two weeks. For Horse, Runway published the character sheets, prompts and production process on X, encouraging users to make their own versions of the film.

"Everyone thinks that AI is reducing things to sameness and a lack of originality, but we actually see it as the opposite," Umpherson says. "These are tools that can amplify unique points of view."

Predicting audiences, with limits

Not every startup is generating footage. Chronicle, co-founded by CEO Aaron Sisto and Scott Greenberg, uses AI to help studios and individual creatives understand how audiences will respond to planned content, and to craft more targeted marketing campaigns and trailers for specific audiences or international markets.

The cautionary tale here is Relativity Media, which pitched data-driven box office prediction in a pre-generative AI world and later went bankrupt. Sisto concedes AI cannot predict how every viewer will respond. "We just don't have the models or the data to do that," he says.

Greenberg, who also co-founded Bento Box Entertainment, the animation studio behind Bob's Burgers, says Chronicle deploys AI agents that watch YouTube channels to understand what content resonates with a creator's fanbase. But he sees a hard limit on algorithmic influence: "You never want to tell a creator what to make. There's nothing worse than an overly opinionated creative executive."

Luma's credits include The Old Stories: Moses, a hybrid generative-AI production backed by Amazon Web Services and starring Oscar-winning actor Ben Kingsley. In March, the company debuted Luma Agents, which can execute creative work across text, images, video and audio on behalf of studios, marketing teams and agencies.

Amit Jain, Luma's CEO and co-founder, says creative roles will expand rather than vanish. A costume designer, in his view, will be responsible not only for what can be physically crafted but also for virtual representations in the digital world. As image and video models improve, "the priority shifts toward doing longer and longer horizon work," Jain says. "The job will still be done by humans, but AI will be doing parts of the task."

Whether audiences reward that division of labor — or keep rewarding practical effects, as the 2026 box office suggests — will determine how much of the billions invested in AI video startups actually converts into Hollywood output.