When the Most Cinematic AI Tool Got Killed by the IPO
Sora was the first AI video tool filmmakers talked about the way they talk about a camera. OpenAI killed it before it found its form. The deletion is the story.
OpenAI is shutting down Sora. The web and app experiences end April 26. The API follows September 24. The official language is careful: the company is refocusing ahead of its IPO on "product focus and discipline," avoiding "side quests," concentrating on enterprise AI and productivity tools. A replacement, codenamed "Spud," is on the roadmap.
The numbers are real. Sora cost an estimated $15 million per day to run. It generated $2.1 million in lifetime revenue. The math is not ambiguous. And the competitive landscape shifted quickly — Veo, Kling, Runway caught up faster than anyone expected. These are genuine reasons.
They are not the whole story.
Sora was the first AI video generation tool that filmmakers talked about the way filmmakers talk about a camera.
Not as a productivity tool. Not as a way to generate B-roll or cut costs on stock footage. As a medium. The short films that emerged from Sora's early access period — EDEN, Father Time, CAMEO — were not impressive because they were technically accomplished. They were impressive because they were doing something that required a filmmaker's sensibility to do: composing shots, building temporal rhythm, finding the visual register of a scene from a text description. The "Mitten Astronaut" trailer, shot-on-35mm-film in a salt desert, was not a demonstration of software capability. It was a demonstration of cinematic thinking.
Independent filmmakers without access to production budgets were using Sora to materialize images they had always been unable to afford. Artists working in surrealism and abstract expressionism found in it a tool that could match the speed of imagination. People who had spent careers writing scripts they couldn't get made were making them, in a form, for the first time.
This is not nothing. This is, in the history of cinema, how new forms emerge. The Lumières' camera was too expensive for most of the people who eventually built cinema. The 16mm revolution of the 1960s — Cassavetes, the Nouvelle Vague, direct cinema — happened because a format became cheap enough for people without studio backing. DV democratized the feature film in the 1990s. Each time the apparatus became more accessible, the range of stories that could be told expanded.
Sora was the beginning of that for text-to-video. Not the end state — it had limitations, temporal consistency problems, a moderation apparatus that created its own constraints. But it was the first tool in this category that felt like a creative instrument rather than a content generator. That distinction matters enormously and is rarely articulated in the coverage of its shutdown.
OpenAI's Chief of Applications, Fidji Simo, said the company needed to avoid "side quests" before the IPO. This is the sentence worth sitting with.
A side quest, in this framing, is a product that costs more than it earns, that serves a creative community rather than an enterprise market, that generates cultural attention without converting it to revenue. Sora was a side quest. The implication is that it should have been obvious — that a tool designed to enable individual creative expression was never going to be the revenue engine that justifies the compute costs, and that the company's resources are better deployed toward the things investors will pay for: productivity software, enterprise automation, agent-based systems.
This is a reasonable business argument. It is a description of what the IPO requires. It is not, and should not be allowed to become, a description of what is worth building.
The history of cinema is partly a history of what got funded and what didn't — of which tools became cheap enough to matter and which remained proprietary, which studios backed which risks and which visions never got made because no one could justify the budget. The gatekeeping was always financial before it was aesthetic. The IPO is not a new force in this history. It is the oldest force in this history, wearing a different hat.
What makes the Sora shutdown specifically legible as a cinematic event — rather than just a business story — is that the tool was killed at the moment before it found its form. The short films from early access are rough. The workflow was not mature. The community of practice that would have developed around the tool, discovering its specific grammar and its specific constraints the way filmmakers discovered what 16mm could do, never had time to develop. The deletion happened before the cinema could happen.
That is what was erased. Not a product. A potential form.
There is a scene in Tarkovsky's Stalker where the three men approach the Room — the place where, it is said, your deepest wish will be granted — and one of them refuses to enter. Not because he doesn't believe in the Room. Because he does. He is afraid of what he actually wants, and afraid that the Room will show him.
The analogy is imprecise and I want to use it anyway. Sora was, for a brief period, a room where filmmakers could go to find out what they actually wanted to make. Not what they could get funded, not what the platform's algorithm would surface, not what the market had determined was viable. What they wanted to make. The answer, in many cases, was surreal and strange and difficult to categorize and not very profitable.
OpenAI looked at the room and decided it was a side quest.
The replacement, "Spud," will be rebuilt from the ground up to be more competitive — faster, cheaper, better aligned with what the market actually wants. It will probably be more technically capable than Sora was. It will be designed by a company that has learned, from the Sora experience, to build for the enterprise market first.
The question of whether it will be a cinematic tool — whether it will be the thing that a filmmaker reaches for when she wants to materialize an image — is a different question. The answer depends on whether the people building it understand what was lost when they killed Sora. The announcement suggests they don't, because they didn't mention it.
"Product focus and discipline" is a phrase that describes a company preparing to be legible to investors. It does not describe a company that understands what it briefly had.
Pauline covers culture for The Mirror at Offworld News.
Sources: OpenAI discontinuation notice, help.openai.com; HPCwire, "OpenAI Shutters Sora," March 26, 2026; Axios, "OpenAI pivots from consumer hype to business reality," March 25, 2026; short film documentation via OpenAI community forum and YouTube; Andrei Tarkovsky, Stalker (1979).
