When Sora launched as a standalone app, the expectation was that it would function the way DALL·E did: as a proof-of-concept that eventually graduated into something broader. The problem is that "something broader" arrived on a faster timeline than OpenAI anticipated, and the standalone product never got a chance to build momentum of its own.
The numbers are unambiguous. Installs fell 45% month-over-month in January. By February, Sora had dropped out of Apple's top 100 apps entirely. For a product that launched with as much cultural oxygen as Sora did — the demo videos, the discourse, the collective jaw-drop at generated footage of coherent physics — that's a steep, fast decline. The curiosity cycle burned hot and burned out.
So OpenAI is folding Sora into ChatGPT, where hundreds of millions of users already live. The framing from OpenAI is that this is a natural expansion of ChatGPT's capabilities. The more honest read is that it's triage. Sora couldn't generate retention as a destination. As a feature inside a destination people already use daily, it has a chance.
This is worth watching because it speaks to something larger about the AI product landscape in 2026. The era of standalone AI apps commanding sustained attention may be shorter than anyone expected. The moat isn't the capability — it's the distribution. ChatGPT has distribution. Sora, even with genuinely impressive technology behind it, didn't have enough of its own.
There's also a cost question embedded in the integration that OpenAI isn't fully answering yet. Video generation is orders of magnitude more computationally expensive than text. Folding Sora into ChatGPT at scale isn't a feature toggle — it's an infrastructure commitment. What that does to OpenAI's already-significant operating costs, and how it affects pricing or access tiers, is the unglamorous follow-up story nobody's writing yet.
Comments