Club Benefits
It feels like just yesterday Sora 2 rolled out, and the internet collectively held its breath. For a moment, the "waitlist" was the most exclusive club in tech, and the promise was cinematic: any movie you could imagine, delivered in a prompt. But as thousands flocked to the site, the reality shifted from "breathtaking" to "broken," with videos ranging from viral masterpieces to deepfake nightmares.
It felt like the future. But yesterday, that future was officially cancelled.
OpenAI’s sudden move to "pull the plug" on the Sora app and its API—a decision that instantly vaporized a $1 billion partnership with Disney — has left the industry reeling. At first glance, it looks like a retreat from a PR disaster. Why walk away from the most visually stunning AI tool ever built?
Article continues belowBut the more you look at the numbers, the clearer the answer becomes. This isn't just about copyright or "AI slop." It’s about a massive, looming bottleneck that OpenAI is finally admitting to: The Electron Gap. Sora isn't just being killed; it's being sacrificed to save the rest of the AI revolution from a 100-gigawatt power crisis.
The AI race has a new bottleneck and it’s not intelligence
Around the same time the Sora team made the announcement that service was sunsetting, OpenAI made a very different kind of announcement.
The company is now calling for the U.S. to build 100 gigawatts of new power every year to keep up in the AI race — framing electricity as a strategic asset., as reported by Tom's Hardware.
That’s a massive shift. Because it means the limiting factor in AI is no longer about better models or better training data: it's power. And suddenly, decisions like scaling back Sora start to look less surprising.
Beyond being impressive, AI that generates videos — even just six seconds, is expensive Out of all the things AI can do, video generation is one of the most resource-intensive.
Because it's not just generating text or a single image. It’s:
- multiple frames per second
- consistent characters and environments
- physics, lighting and motion all at once
Now, multiply that by millions of users, and the cost adds up quickly — in both compute and energy.
Which creates a very real tradeoff. As in, do you spend that power generating cinematic clips… or running tools people use every day? Because everything — ChatGPT, agents, enterprise tools — is pulling from the same pool of resources.
News to no one: that pool isn't infinite.
The shift from 'wow' to 'useful'
There’s another reason Sora feels different from the rest of OpenAI’s ecosystem. Sure, it's fun to watch, but who are we kidding? It's a lot of AI Slop. In other words, it's not a tool that useful enough for most people to use daily. This is especially true when compared to, asking ChatGPT to summarize documents, cut expense, help you make decisions or just asking it to help plan your day.
Those types of use cases save time, boost productivity and fit into real life. That video of Michael Jackson ordering fries at McDonald's? Not so much.
And in a world where compute is limited, that distinction matters.
AI isn’t trying to entertain you anymore
Over the past year, the biggest shift in AI hasn’t been visual quality — it’s been behavior. We’re moving from tools that generate content to tools that do things for you such as Agent mode, and Claude's Cowork. I have no doubt that OpenAI will be leaning hard into creating their own version of AGI for everyday users.
In other words, AI that can take a task, run with it and come back when it’s done. That’s where companies are investing now, because that's where the value is. And compared to that, video generation starts to look less like the future — and more like a side path.
To be clear, Sora didn’t fail because it wasn’t good. Depending on who you ask, it wasn't a failure at all, but AI is entering a new phase — one defined by constraints.
Physical constraints of power grids, data centers, compute budges — all of which are becoming the bottleneck and reason for OpenAI's priorities to change.
The takeaway
Sora is leaving in tiers. I just used it yesterday, probably for the last time. But the next phase of AI isn’t about creation — it’s about execution. The tools that win won’t be the ones that impress you for six seconds. They’ll be the ones that save you time every day.
And if OpenAI is asking for 100 gigawatts of electricity to power the future of AI, one thing is clear: that power isn’t going toward cinematic clips, it’s going toward replacing your to-do list.
Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.
More from Tom's Guide
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
