OpenAI killed SORA, its superstar app, and it tells you everything about where AI video won’t make any money.
As OpenAI walks away from viral video, the future of AI shifts from spectacle to invisible tools that actually ship work.
Yesterday when I wrote ‘that while everyone was waiting for the “next Sora moment,” AI was busy eating the creator workflow one tool at a time’, I didn’t realize today’s headline would be: there is no next Sora moment -at least not from OpenAI. 😏
What just happened?
OpenAI is shutting down Sora. the app, the API, and the video features wired into ChatGPT, just months after its big launch. The official line is “refocusing” and “goodbye Sora,” but practically it means one of the most hyped text‑to‑video products is being taken off the table entirely.
For a brief window, Sora was the demo:
Hollywood‑style shorts from a sentence, a headline deal with a major studio, and a sense that text‑to‑video had finally “arrived.” Now the consumer and dev surface are being retired, while OpenAI keeps the underlying video tech in‑house for things like simulation and robotics.
Why Sora died?
Underneath the PR, three reasons keep coming up in reporting and forums:
Cost:
High‑quality video gen is insanely GPU‑hungry. Long, HD clips blow up compute in a way text and images don’t, and most users were never going to pay anywhere close to what they cost to serve.
Risk:
Sora was already attracting heat over deepfakes, dead celebrities, and copyrighted characters wandering into “fan” clips, right as regulators and Hollywood were waking up to AI video.
Strategy:
and This might come as a surprise to video developers but not to founders and builders who’ve been keeping an eye on OpenAI/ that Ahead of an IPO, OpenAI needs to look like a disciplined, enterprise‑first company, not a lab running loss‑making side quests. Video was the most expensive, and the least obviously monetizable one of their side projects and it had to die.
So Sora isn’t being shut down because the model was bad.
It’s being shut down because, in a world where every GPU hour now has to defend itself to a CFO, “free cinematic clips for the internet” loses to code copilots, agents, and enterprise tools. that make me question, how much are you using Veo3 and the NotebookLM cinematic video feature? don’t worry though, Google got lot of money.
The Disney “rug pull” and trust!
One very intersting subplot here is what this looks like to big partners.
A major studio had a multi‑year, billion‑dollar‑scale collaboration around Sora lined up, with teams already building on it, and then learned support was being discontinued more or less at the same time as everyone else.
The investment hadn’t fully landed yet, but the optics are brutal.
if Sora can disappear overnight for a top‑tier media company, it can disappear overnight for you. and,
That’s the real platform‑risk story for founders and studios.
and this is a lesson for us. It’s not just that “one cool app is going away,” it’s that any shiny AI product, especially the demo‑driven ones, can be switched off by a spreadsheet somewhere. Today it’s video. Tomorrow it could be the API you quietly wired into three of your workflows. 😉
How people are reacting?
Across LinkedIn, Twitter, and Reddit, the reaction is surprisingly aligned. some examples here -
LinkedIn:
this is financial discipline. Impressive tech that doesn’t map to clear, repeatable spend gets cut so compute can be reallocated to enterprise tools, agents, and dev platforms.
Twitter/X:
“end of GPU flex culture.” Sora is being cast as the ultimate demo, amazing to watch, terrible economics, and people assume those GPUs are already earmarked for the next foundation model.
Reddit:
“cool, but I never really needed it.” Threads are blunt about using Sora for a bit, getting wowed, then drifting back to their normal tools, while pointing out that the cost curve was obviously upside‑down.
What I find the most interesting is that, No one’s arguing the tech wasn’t jaw‑dropping. They’re arguing it never crossed the gap from spectacle to necessity.
I dont’ want to do a OpenAI business model deep dive here, but.
I have to mention this.
OpenAI’s business? ChatGPT subscriptions, enterprise licences, and API usage with custom deals on top, and they do now have revenue 10x billons annulized, ( lets not think of loss for a minute) but it’s sitting on top of a capex plan that involves tens (maybe hundreds) of billions for chips and data centres over the next few years.
In that context, Sora is the first visible casualty of a new rule:
every product has to justify its share of the GPU bill.
I would come back to the whys and other deeper questions like return on compute, margins, durability soon on the Intelligent founder AI.
[ In this newsletter you get sharp, unfiltered short essays; for full‑length, deep‑dive analysis on AI, subscribe to our companion publication, Intelligent Founder AI. ]
So finally what now for video and content creators and competitors?
Sora going away doesn’t kill text‑to‑video. It just forces everyone to grow up.
Other players (Runway, Pika, Luma, Google’s Veo, plus the China stack) suddenly have more room. BUT also tougher questions from investors and customers about their own economics.
If you’re a founder or studio, the takeaway is: treat text‑to‑video as a component, not the whole product. Don’t build a business that only works if one proprietary API stays cheap and immortal. I think I said this before as well, in terms of models and other products non AI including. the lesson is same. Use multiple providers, and keep video gen at the edges of your workflow ie. pre‑viz, drafts, B‑roll and not as the only way you ship.
And most importantly, this all reinforces the thesis from yesterday’s post: the real action is in the “boring” parts of the creator stack. Masking, mattes, clean‑up, localisation, asset search, compliance jobs where AI is quietly making editors 5–10× faster and the cost/ROI maths actually works.
If Sora was peak AI spectacle, the endless synthetic clips with fuzzy business value? the next wave of video AI will look more like invisible plumbing:
agents and tools that sit inside Premiere, Resolve, and Notion, automate the grunt work, and never trend on X.