OpenAI killed Sora on April 26, 2026. The text-to-video app that was supposed to anchor the company’s entertainment-industry push lasted six months in market before the web and mobile experiences went dark, with the API set to follow on September 24. The collateral damage is the bigger story: Disney has walked away from a planned $1 billion investment in OpenAI, learning about the OpenAI Sora shutdown from the public announcement less than an hour before everyone else. For an industry that has been telling itself AI video was the next inevitable platform, this is a hard reset.
What’s actually new
The substance of the shutdown is concrete and the timeline is documented.
- Sora web and mobile apps discontinued April 26, 2026. Users with active subscriptions received pro-rated refunds. Generated videos remained downloadable through May 31, 2026 before being purged.
- Sora API discontinuation set for September 24, 2026. Developers and enterprise customers building on Sora’s API have approximately four months to migrate to an alternative platform.
- User collapse drove the decision. Sora peaked at roughly 1 million users globally in its first weeks, then fell below 500,000 by mid-2026. Daily active engagement reportedly dropped to a fraction of even those reduced totals.
- Compute economics never worked. The widely cited figure inside OpenAI is that Sora was burning more than $1 million per day in compute costs against negligible revenue. Some industry reports place the daily burn substantially higher when fully-loaded infrastructure costs are included. The unit economics never closed.
- Disney walked from the planned $1 billion investment. The deal would have given Disney a strategic stake in OpenAI alongside content partnerships. Disney’s leadership reportedly learned of the Sora shutdown roughly an hour before the public announcement, and pulled its investment commitment within days.
- No official replacement. OpenAI has not announced a successor product. The company’s video-generation research continues; what it ships next is undecided. The implication: the next video product, if it exists, will arrive on a different operating model than Sora’s.
Why it matters
- AI video economics are harder than the rest of the AI stack. Image generation is profitable at consumer prices. Text generation is profitable at consumer prices. Video is not — yet. The OpenAI Sora shutdown publicly confirms what insiders had been arguing for months: at current model and infrastructure costs, video generation cannot be priced at consumer accessibility without burning capital.
- The Disney withdrawal recalibrates AI-entertainment partnerships. A $1 billion media-company investment in a frontier AI lab evaporating because of one product shutdown changes how every studio negotiates with every AI vendor for the next 18 months.
- The market opportunity shifts to Runway, Pika, and HeyGen. The companies that have built video generation as their core business — not as one product among many — pick up the API customers, the prosumer subscribers, and the enterprise relationships Sora was incubating.
- OpenAI’s resource focus tightens. The implicit message: GPT-class reasoning models, agentic platforms, and enterprise deployment win. Consumer-creative experiments lose. Expect more “discontinuation” headlines for OpenAI side-projects as the company funnels compute toward higher-margin work.
- Trust impact on developers. Anyone who built on Sora’s API and got six months of notice before discontinuation has a vivid lesson in AI-platform risk. Multi-vendor backend strategies become more attractive.
- The “consumer AI app” pattern is in question. Sora was OpenAI’s experiment in shipping a consumer-facing creative app distinct from the ChatGPT product. The shutdown suggests that pattern is harder to make work than the ChatGPT-centric product strategy that drives OpenAI’s revenue today.
How to use it today
If you were using Sora, building on it, or planning to integrate, this is the migration sequence that works in May 2026.
- Inventory your Sora dependencies before September 24. Any production workflow, any API customer, any in-flight prototype. List them now while there is time to migrate.
# Example: audit your Sora API usage grep -rE "api\.openai\.com/v1/video|sora\." ./your-codebase --include="*.py" --include="*.ts" --include="*.js" # Or use OpenAI's usage dashboard curl -X GET "https://api.openai.com/v1/usage?start_date=2026-01-01&end_date=2026-05-10" \ -H "Authorization: Bearer $OPENAI_API_KEY" | jq '.data[] | select(.model | contains("sora"))' - Download any Sora-generated videos you want to retain. The May 31 deadline for downloadable retention has passed for general users; check OpenAI’s account dashboard for any retention paths still active. Anything in your local files is safe.
- Migrate to a working alternative. Pick the one that matches your workflow. See the comparison table in the next section for the trade-offs.
# Migrating from Sora API to Runway Gen-4 API import os import httpx RUNWAY_API_KEY = os.environ["RUNWAY_API_KEY"] async def generate_video(prompt: str, duration_seconds: int = 5): async with httpx.AsyncClient(timeout=300) as client: r = await client.post( "https://api.runwayml.com/v1/generate", headers={"Authorization": f"Bearer {RUNWAY_API_KEY}"}, json={ "model": "gen-4-turbo", "prompt": prompt, "duration": duration_seconds, "aspect_ratio": "16:9", "resolution": "1080p", }, ) return r.json() - Hedge with multi-provider abstraction. If you are still building video-generation features, do not commit to a single vendor. The Sora shutdown demonstrates that even category-defining products from category-defining companies can disappear with limited notice. Wrap video generation behind your own provider-agnostic interface.
- Re-price your video features. If your product passed Sora’s costs through to customers, recheck the pricing math on the replacement provider. Runway, Pika, and HeyGen all have different pricing structures; some are more expensive per-second of output than Sora was.
- Communicate proactively with downstream users. If your customers depended on Sora-generated content through your product, tell them what is changing and what stays the same. The communication does not need to be alarmist; it does need to be early.
How it compares
The video-generation landscape after Sora’s exit:
| Platform | Strengths | Pricing | Best fit |
|---|---|---|---|
| Runway Gen-4 | Highest quality among public APIs; broad creative tooling; agency relationships | $0.05-$0.12/sec generated | Production studios, advertising, agencies |
| Pika 2.0 | Creative versatility; community of remixers; consumer-friendly UX | $0.04-$0.08/sec | Social-first creators, indie filmmakers |
| HeyGen | AI avatars + scripted video; enterprise sales workflows | $0.10-$0.25/sec depending on avatar | Sales, training, internal communications |
| Synthesia | Avatar-driven corporate video; multilingual; compliance-friendly | $22-$67/user/month plans | Corporate L&D, customer education |
| Luma Dream Machine | Fast generation; mobile-friendly; cinematic outputs | $0.02-$0.06/sec | Mobile creators, social content |
| Kling AI (Kuaishou) | Strong character consistency; available outside Western platforms | $0.03-$0.07/sec | Global content, non-US markets |
The replacement decision depends on workflow more than capability. For Sora’s prosumer-creator base, Pika and Luma are the natural landing points. For Sora’s enterprise API customers, Runway is the dominant replacement. For the entertainment-industry partnerships that Sora was supposed to anchor — those are mostly migrating to a mix of Runway for production and HeyGen for promotional video.
What’s next
Three threads to watch in the months after the shutdown.
OpenAI’s next video play. The company has not announced what comes after Sora. Industry speculation centers on a more selective, enterprise-focused offering that ships when the compute economics work — likely on next-generation hardware (Blackwell Ultra at scale) and possibly bundled into the broader ChatGPT product rather than as a standalone app. Expect a quieter launch than Sora’s December 2025 debut.
Disney’s AI strategy reset. The pulled $1 billion investment does not mean Disney is exiting AI partnerships — it means Disney is reconsidering how to structure them. Expect new partnerships with multiple AI labs at smaller individual investment levels, or strategic deals with Runway, Pika, or other dedicated video labs. The “single AI partner” model that Disney-OpenAI represented may not return.
The AI video economics cycle. Sora’s shutdown is the first major case study in AI video economics not closing. It will not be the last. Expect 2026 to see additional shutdowns or pivots from companies that launched video generation as a feature without honest cost analysis. The survivors will be the labs that built video generation as their full-time business with operating model alignment to those economics.
Frequently Asked Questions
When did OpenAI shut down Sora?
The Sora web and mobile apps were discontinued on April 26, 2026. The API remains operational until September 24, 2026, giving developers and enterprise customers approximately four months to migrate to alternative providers.
Why did OpenAI shut down Sora?
OpenAI did not give a single official reason. The factors that have been cited publicly and inside the company: rapid user churn (from a peak around 1 million to below 500,000), unsustainable compute costs (reportedly $1 million+ per day in operating expenses against minimal revenue), strategic refocus on core enterprise products, and pressure on engineering and infrastructure resources from higher-priority projects including GPT-5.5 and agentic platforms.
What happened to Disney’s $1 billion investment?
Disney withdrew the planned investment after learning about the Sora shutdown approximately one hour before the public announcement. The investment had been linked to a broader strategic partnership including content licensing and AI-enabled production workflows. Disney has not publicly committed to a replacement partnership at the same scale.
What can I use instead of Sora?
For prosumer creative work: Pika 2.0 and Luma Dream Machine are the closest UX replacements. For enterprise API use: Runway Gen-4 is the most capable and broadly adopted. For avatar-based corporate video: HeyGen and Synthesia are mature options. For social-first creators outside Western platforms: Kling AI from Kuaishou. The right choice depends on your workflow and pricing tolerance.
Will OpenAI release another video model?
Probably, but on a different operating model than Sora’s. Statements from OpenAI leadership and reporting around the shutdown suggest a future video offering will arrive when compute economics improve — likely on next-generation NVIDIA Blackwell Ultra hardware at scale and possibly bundled into the broader ChatGPT product rather than as a standalone app. No timeline has been confirmed.
Should I avoid building on OpenAI APIs after the Sora shutdown?
No, but the prudent posture is provider-agnostic. The Sora shutdown demonstrates that even major products from major AI labs can be discontinued on relatively short notice. Build your video, image, and text generation capabilities behind your own abstraction layer so that swapping providers — including switching away from OpenAI — is a configuration change, not a code rewrite. The core OpenAI products (GPT models, the Assistants API, Whisper) remain well-supported; the lesson is about portfolio resilience, not vendor avoidance.