The future of entertainment was supposed to be a major investment. Instead, it proved too expensive to run. OpenAI, artificial intelligence’s leading firm, confirmed it was shutting down Sora, its consumer video generator. The decision vaporised a landmark partnership with Disney, announced just months earlier. The deal did not fail on a lack of imagination or a flaw in the technology. It failed on the grim reality of a balance-sheet.
The pact, unveiled in 2025, was meant to be a coronation for generative video. Disney, a titan of intellectual property, agreed to license many of its most iconic characters for use in Sora. For a subscription, users could conjure new shorts starring well-known figures. A Disney executive declared the deal would “unlock new possibilities in imaginative storytelling”. For OpenAI, it was the ultimate validation from the creative establishment, backed by a pledged significant equity investment from Disney.
But the unit economics were ruinous. Sora was considered the gold standard of its kind, yet its business model was unsustainable. Generating a single video took several minutes of intense computation. To cover this, OpenAI charged users a high monthly fee. Competing services from firms like Pika Labs, while perhaps less powerful, were operating profitably at a much lower price, with faster generation times. With a projected high cash burn in the coming years and a potential IPO on the horizon, OpenAI made a strategic choice. It would pivot from costly consumer-facing products to more lucrative enterprise tools.
The plug was pulled before any money changed hands. “We’re saying goodbye to Sora,” OpenAI announced in a statement. “What you made with Sora mattered, and we know this news is disappointing.” Disney, left at the altar, was diplomatic. “We respect OpenAI’s decision to exit the video generation business,” a spokesperson said, confirming the company would not proceed with its investment.
For a studio, adopting a third-party AI provider presents analogous financial hurdles and operational risks, a concern echoed in reports on third-party risk from the Bank of England and the BIS.
The entertainment industry now faces a sobering new reality. The initial euphoria around text-to-video models, which promised to democratise filmmaking and slash production costs, has been replaced by a stark understanding of their financial burden. The industry’s AI future is not dead. It is being recalibrated from a public spectacle to a back-office utility. The question for studios is no longer what AI can create, but what they can afford for it to create.
This forces a sharper “build versus buy” dilemma. With the premier off-the-shelf option gone, studios must choose a new path. They could invest heavily in proprietary AI systems, a high-stakes gamble far from their core competence of making films. Or they can partner with smaller, nimbler firms like Runway or Pika Labs. Yet these challengers, now free from Sora’s shadow, face the same unforgiving economics if they chase the mass market. The cost of integrating such complex systems is also a deterrent. The banking industry offers a cautionary tale. Modernising core banking platforms or integrating legacy systems routinely costs “tens of millions” of dollars and takes “months” to complete. For a studio, adopting a third-party AI provider presents analogous financial hurdles and operational risks, a concern echoed in reports on third-party risk from the Bank of England and the BIS.
To be sure, the dream of AI-powered mass entertainment has not vanished. Some will argue that Sora’s failure is a temporary setback for one company’s business model, not a final verdict on the technology itself. Computing costs will eventually fall. New, more efficient models will emerge. The commercial prize of letting millions of fans co-create stories with their favourite characters is too great for media giants to abandon forever. This was just the wrong model at the wrong time.
Yet this view mistakes a strategic retreat for a tactical one. The problem was not Sora’s code, but its cost. OpenAI, a company with more capital, data and talent than any rival on Earth, could not make the numbers work. This signals that the basic economics of generating high-fidelity video for a mass consumer audience are, for the foreseeable future, broken. Any competitor will be fighting the same laws of financial gravity. Profitability lies not in mass-market video toys, but in targeted, high-value professional tools where the cost per minute can be justified by production savings.
The focus of AI in entertainment will therefore shift from the audience to the artist—or at least, to the production pipeline. The technology will be integrated into pre-production to generate storyboards, into visual-effects to render complex scenes and into marketing to create endless variations of trailers. In these controlled, professional settings, costs can be managed and returns on investment measured. It is less a revolution in public creativity than an evolution in industrial efficiency.
This is the hard lesson from the deal’s collapse. The initial promise of AI in the entertainment industry was that of a magical new paintbrush for every fan. The reality is that it is a powerful, and expensive, new tool for the studio assembly line. The magic is real, but it will be happening behind a curtain of immense cost, wielded by professionals, not the public. The industry has not found a new form of entertainment; it has found a new form of automation.
