Sora Shutdown — Why OpenAI Killed Its Own Video Generator

Why OpenAI Killed Its Own Video Generator

Imagine building a product where every 10 seconds of usage costs you $130. Then discovering your total lifetime revenue is just $2.1 million — while you’ve been spending $15 million per day to keep it running. That’s the real story of Sora, OpenAI’s AI video generation model that was officially shut down less than six months after its public launch.

On March 24, 2026, OpenAI announced it was discontinuing Sora. The web and mobile apps closed on April 26, and the API will shut down by September 24, 2026. What was supposed to be a revolutionary product became the biggest AI financial disaster of the year.

What Was Sora and How Did It Start?

Sora was a text-to-video AI model. You’d write a text prompt, and the AI would generate a video for you. The first version launched in limited access for ChatGPT Plus and Pro users in December 2024. Then Sora 2 arrived in September 2025 with dedicated iOS and Android apps.

The early demos were genuinely stunning — videos of Tokyo streets, fantasy creatures, and cinematic scenes that shocked the industry. Media hype went through the roof. But there’s a deep chasm between “impressive demo” and “sustainable product.”

Analogy

Sora was like a luxury restaurant serving every meal at a loss. The food was extraordinary, but each plate cost $130 to make while customers only paid $2. How long can you keep that going?

The Terrifying Numbers: $15 Million Per Day in Losses

Let’s put the figures side by side to understand why continuing was impossible:

Cost per 10-second video: approximately $130
Daily compute costs: up to $15 million
Estimated annual cost: around $5.4 billion
Total lifetime revenue: just $2.1 million

Yes, you read that right. $5.4 billion in costs versus $2.1 million in revenue. That’s roughly $2,500 spent for every $1 earned. Even for a company like OpenAI that has raised billions in funding, this was impossible to justify.

Key Takeaway

AI video generation is exponentially more expensive than text or image generation. A text prompt in ChatGPT costs a few cents, but a 10-second video costs $130. This thousand-fold difference shows why AI video hasn’t reached economic maturity yet.

Users Walked Away Too

It wasn’t just costs. Users left after the initial excitement faded. App downloads dropped from 3.3 million in November 2025 to 1.1 million by February 2026 — a 66% decline in three months. Active users fell from about one million to under 500,000.

Why? Because Sora’s videos still had significant issues. Object physics were unreliable, faces would sometimes distort, and videos longer than 10 seconds were practically unusable. For professional content creators, it was a novelty, not a tool.

Fidji Simo: No More “Side Quests”

Fidji Simo, OpenAI’s applications CEO, told employees the company could no longer afford “side quests.” Sam Altman personally made the call to shut Sora down and redirected compute resources to higher-priority products — coding tools and enterprise services.

The decision even derailed a reported $1 billion partnership with Walt Disney. According to The Wall Street Journal, Disney had planned to integrate content from its franchises with Sora but was informed shortly before the public announcement. No money had changed hands.

Business Lesson

Even OpenAI, with billions in funding, can’t sustain a money-losing product indefinitely. The key lesson: “exciting technology” doesn’t automatically equal “viable business model.” Before investing in any technology, first make sure it actually serves your needs.

The AI Video Market After Sora

Sora’s shutdown doesn’t mean AI video is dead. Competitors are filling the gap:

Google Veo 3.1: Currently the best overall quality. It offers 4K output, native audio, and even a free tier — something Sora never had.
Runway Gen-4.5: The top choice for professionals in advertising and narrative content, with the best camera motion control available.
Kling 3.0 (by Kuaishou, China): More realistic physics at roughly 40% less cost than Sora. Ideal for high-volume social media content production.

The market has now split into four tiers: quality-first (Runway), cost-efficient (Kling), ecosystem-integrated (Google Veo), and open-source (Seedance). Professional studios are using combinations of multiple tools.

Lessons from Sora’s Failure

1. A demo is not a product. Sora’s demos were spectacular. But a real product must be scalable, economical, and reliable. Many AI projects look incredible in demo stage but fall apart at scale.

2. Compute costs determine everything. Until the cost of AI video generation drops dramatically, this technology won’t become a consumer product. Right now, it’s a specialized tool for studios.

3. Even giants fail. OpenAI, valued at over $150 billion with the world’s best engineering teams, couldn’t save Sora. If you’re planning to use AI in your business, learn the common mistakes so you don’t fall into the same trap.

4. Focus beats diversification. By shutting down Sora, OpenAI concentrated its resources on ChatGPT, coding tools, and enterprise services. Sometimes the best business decision is saying “no” to an attractive but unsustainable product.

Bottom Line

AI video isn’t dead — it just hasn’t reached the “anyone can use it for free” stage yet. Competitors like Google and Runway are continuing down the path. But Sora’s story proves that even the most advanced technology is doomed without a sound business model.

Sources: TechCrunch | OpenAI Help Center | TechXplore