Eighteen months ago, Sora was the most-hyped product in AI video. As of April 26, 2026, it's gone. OpenAI quietly discontinued the Sora web and app experiences, with the developer API scheduled to follow on September 24, 2026. For a model that once topped a million users and dominated launch-day headlines, it's a remarkably abrupt ending — and a useful lesson about the economics underneath the generative-video boom.
From launch hype to shutdown notice
Sora arrived in two acts. The first generation launched to ChatGPT Plus and Pro subscribers in late 2024, generating short, cinematic clips from text prompts with a then-unmatched grasp of scene continuity and physics. Sora 2 followed at the end of September 2025, adding synchronized audio, longer clips, cameos, and a standalone consumer app that briefly shot up the download charts.
Then the curve bent the wrong way. Reporting around the shutdown points to a worldwide user base that peaked near a million before sliding below 500,000, against an operating cost estimated at roughly $1 million per day. Video generation is brutally compute-hungry, and unlike text, every second of output burns GPU time that's hard to recoup at consumer price points. OpenAI's own notice didn't give a single reason, but the pattern is clear: compute shortages, cost pressure, and a deliberate refocus on core enterprise and ChatGPT products.
Why video is harder to sustain than text
The Sora story isn't really about one product failing — it's about the unit economics of AI video. A text response costs fractions of a cent. A few seconds of 1080p video can cost orders of magnitude more, and users generate it speculatively, tossing dozens of takes to get one good clip. That dynamic makes flat-rate consumer subscriptions structurally difficult: the heaviest users are also the least profitable, and the "wow" factor that drives signups doesn't translate into durable daily habits the way chat does.
There's also a content-moderation tax. A consumer video app that can render realistic people invites deepfake, likeness, and copyright headaches at a scale that pure-text tools never face. Maintaining that safety apparatus is expensive in both engineering and reputation — another line item that's easier to justify for a flagship chat product than for a second-tier video app with declining usage.
Where AI video creators should look next
If you built a workflow around Sora, the good news is that the category is healthier than ever — just more distributed. A few strong options are worth migrating to now rather than waiting for the API sunset:
- **Runway** remains the most production-oriented choice, with Gen-3-class generation plus a genuine editing suite (motion tracking, background removal) that creators actually ship with.
- **Kling AI** offers some of the longest clips on the market — up to two minutes — with strong physics simulation and a usable free tier, making it a natural landing spot for narrative work.
- **Pika** leans into playful, social-first effects and targeted region edits, ideal for short-form content rather than cinematic sequences.
- **Google's Veo** line and several open models round out the field for developers who want API access without single-vendor risk.
The practical takeaway: don't anchor a business-critical pipeline to a single hosted video model. Sora's shutdown is the second high-profile reminder in a year (after the Codeium-to-Windsurf-to-Cognition saga) that even well-funded AI products can change hands or vanish on short notice. Keep your prompts and source assets portable, and treat any single generator as replaceable.
The bigger signal
Sora's exit doesn't mean AI video is cooling off — adoption across Runway, Kling, and others is still climbing. It means the market is sorting out who can actually afford to run these models at scale. Expect more usage-based pricing, more credits-per-generation meters, and fewer all-you-can-eat consumer plans. The magic of typing a sentence and getting a movie clip isn't going away. The free lunch is.
**The one-line takeaway:** Sora proved text-to-video could captivate millions; its shutdown proves that captivating millions isn't the same as sustaining them — so build your video workflow on tools you can swap out.