OpenAI launched Sora in September with enormous hype. An AI video generation app. Hit number one in the iOS Photo and Video category on day one. Sparked a billion-dollar partnership with Disney. Six months later, they killed it.
The reason OpenAI gave is resources. They said they cannot do everything at once, and the compute chips powering Sora are more valuable running coding, reasoning, and text generation. In plain English: AI video generation costs a fortune to run, and the money is in text-based AI work. The app peaked at about 3.3 million downloads in November and dropped to 1.1 million by February. That is a two-thirds decline in three months.
The Disney deal is dead too. Disney had planned to invest a billion dollars in OpenAI partly around this technology. A Disney spokesperson said they "respect OpenAI's decision to exit the video generation business." Which is the politest way of saying the whole thing fell apart.
This was not a startup experiment that failed
This is the part that matters if you run a business and you are building workflows around AI tools. Sora was not some scrappy side project from a company nobody has heard of. It was OpenAI's flagship creative tool, backed by the biggest entertainment company on the planet, and it still got pulled. If that can disappear in six months, anything can.
I think a lot of people assume that the big players are safe bets. That if you build your process around ChatGPT or Claude or Gemini, those things will be there next year. And they probably will be, bear with me here. But specific features, specific tools, specific integrations within those platforms? Those are a different story entirely. Companies are making hard choices about where to invest compute, and features that do not generate enough return get cut. Even popular ones.
Building around capabilities instead of products
When I set up AI workflows inside client businesses, we build around what AI can do, not around a specific product. There is a reason for that, and Sora is a perfect example of why.
Say you had built your entire content production pipeline around Sora. Your team had learned the tool. Your process documentation referenced it. Your timelines assumed it existed. Now it is gone, and you are starting from scratch. The workflow knowledge is still useful but the specific implementation is worthless.
If instead you had built a content production workflow that used AI video generation as a capability, with the understanding that the specific tool might change, you would be in a completely different position today. You would swap in a different tool and the process keeps running. The SOPs get a minor update. The team adapts in a day because they understand the workflow, not just the software.
That is the difference between building on a product and building on a capability. One breaks when the vendor makes a decision. The other bends.
What to actually do about this
If you are using AI tools in your business right now, which I hope you are, have a think about how dependent you are on any single one. Could you swap it out next week and keep going? If the answer is no, that is your vulnerability.
The teams I work with document their workflows in terms of what happens at each step, not which tool does it. The tool is a detail that can change. The process is the thing that stays. When Sora users woke up this week to find their tool was dead, the ones who had built properly lost a tool. The ones who had not lost a process.
I help businesses build AI workflows that survive exactly this kind of thing. If you want your team's processes to outlast any single tool decision, here is how the fractional AI engagement works.
