M.G. Siegler • #openai • Mar 25, 2026

So Long, Sora

An OpenAI Code Red? More like a Code Dead...

Well, I'd like to say it was fun while it lasted. But the reality is that it barely lasted. I speak, of course, of Sora. And specifically the productized version of OpenAI's video models which launched to much fanfare just a few months ago and quickly rocketed to the top of the App Store. OpenAI brought the axe down so quickly that it apparently surprised both their key partner Disney and even the team internally working on Sora. Even by OpenAI standards, this seems pretty wild.

Of course this also wasn't totally out of left field. A couple weeks ago, a report suggested that Sora was over as a stand-alone app and service. And that seemingly made sense. After the aforementioned initial surge, Sora slowly slid down the download charts and usage seemed to be dwindling over time. As it turns out, much like the Studio Ghibli situation before it, it was a mere viral moment in time. OpenAI is great at creating these, but they're fleeting. That's a problem when you've built a whole service around one.

Sora was the most fun I recalled having with a new app in a while, but I also noted that the key probably wasn't the TikTok-style network, but instead how it eased the creation process for video. I thought individual sharing of such videos was more interesting than pushing them out publicly because clearly most of the novelty was around the insertion of people you knew in the videos. Celebrities inevitably tried to tap into this, but it all felt a bit insincere at best, and cash-grabby at worst.

So yes, it made sense to shift the focus from the app to bring such tools into ChatGPT itself. In particular as the service tried to do more social/collaborative things there – bringing more chat into ChatGPT, as it were.

But a funny thing happened on the way to the roll-up: OpenAI clearly started to freak the fuck out about the traction Anthropic was suddenly seeing.

That mixed with the building push to go public is undoubtedly a big part of what shifted a move to bundle Sora to a move to murder Sora. This is all related. Anthropic is also said to be making moves to go public. And because of their position in the market – in the enterprise market in particular – and their path to profits looking much clearer, this was always going to be a problem for OpenAI. But the moment Claude Code – and tangentially, Claude Cowork – is having probably pushed it all over the edge.

Last week came news that "side quests" would no longer be a part of OpenAI's path. And was there any bigger side quest than Sora? Again, you'd think they could just kill the app and stand-alone service, bringing it into ChatGPT itself – as seemed to be the plan – but the reality there remained just how expensive it was to actually create such videos. OpenAI probably ran then re-ran the numbers. There was just no way to justify it with a straight face if they were truly hoping to paint a path to profitability picture to Wall Street any time soon.

Fine. No fun, but we get it. OpenAI has bigger GPUs to fry.

But what's really wild is shafting Disney in the process. I mean, they're without question the most important entertainment and content company in the world. And OpenAI seemingly did them dirty. The fact that Disney is no longer looking to do the major investment which sure seemed like a good idea to them – checks notes – just three months ago, seems to say a lot. You have to imagine OpenAI would still welcome such an investment. They need any and all money! Even without Sora, you have to imagine there was a path to future partnerships here. But no. Not the way OpenAI executed this. Ouch.

And my god how incredibly awkward for Josh D'Amaro, just one week into his new role as CEO of Disney. While Bob Iger put his name on this deal and subsequent PR around it, all indications were that D'Amaro was the one who really pushed for it. And it played up to technology prowess being a key point of his ascension and plan. This plus the news that Fortnite maker (and key Disney partner/investment) Epic is doing massive layoffs this week.. Again, ouch.

Anyway, back to Sora. This sucks. Not because Sora was particularly useful – though I maintain it was good fun, even now opening it, I get sucked into a rabbit hole that I fill with laughter. I wasn't getting sucked in nearly enough for it to make sense stand-alone, but certainly the technology was compelling! Which is why Google, Meta, and others have been working on video models as well, obviously. It looked like OpenAI had played them with the Disney deal, until they realized that they played themselves.

Was it an IP minefield? No doubt. Did it raise some very real ethical questions about raising the dead to have them do dumb shit in viral videos? For sure. But it was also pretty fun. And very funny. And wonderfully simple to use.

All this sucks because it points to a world in which OpenAI is definitely less product-focused going forward and much more about drilling down on the business. That's undoubtedly the right move for the company given everything going on, but that doesn't make it the right move for me, the end user.

Rest in peace, sweet viral prince. We hardly knew ye. I'm so Sora for our loss.

👇

Previously, on Spyglass...