Last Updated on July 7, 2026 by Eytan Bijaoui
One million dollars. Every single day. For six months.
That’s what OpenAI spent keeping Sora alive. And on March 24, Sam Altman pulled the plug anyway. Not because the technology failed. The technology was stunning. But because fewer than 500,000 people in the entire world could be bothered to use it regularly. A product so expensive to run that each 10-second video clip cost $1.30 to generate, and so few people cared that the user count was falling, not growing.
This is OpenAI. The company that just raised $122 billion in a single quarter. The company valued at $852 billion. The company with more cash, more talent, and more GPU power than most countries. And they couldn’t make people want an AI video tool.
If that doesn’t terrify you as a founder, you’re not paying attention.
The Disney Disaster
Here’s the part of this story that keeps me up at night.
In December, Disney announced a massive partnership with OpenAI. Their beloved characters would come to Sora. A three-year deal. And Disney committed to investing $1 billion in OpenAI as part of the agreement.
Three months later, OpenAI killed the product. Disney found out less than one hour before the public announcement. Not a week. Not a day. Less than sixty minutes.
A billion-dollar deal, dead. And Disney’s executives were basically the last to know.
I’ve seen startups handle partner communications badly. I’ve seen founders ghost investors after a pivot. But telling the most powerful entertainment company on Earth that their billion-dollar partnership is over, with less notice than it takes to watch a Disney movie? That’s a new level of “we just don’t have a choice anymore.”
The deal collapsed. No money changed hands.
Why It Actually Died
Let’s be clear about something. Sora didn’t die because the technology was bad. The videos it generated were genuinely impressive. The visual quality was miles ahead of where AI video was just two years ago. The technology worked.
The business didn’t.
Here’s what happened. After the initial hype cycle, users showed up, played around, made a few clips, and then left. The worldwide user count peaked around one million, then dropped below 500,000 and kept falling. Nobody was building Sora into their workflows. Nobody was replacing their video production with it. Nobody was paying for it in a way that justified $1 million in daily compute costs.
And while OpenAI’s team was pouring resources into making Sora work, something else was happening. Anthropic was quietly winning the customers that actually pay money. Claude Code was eating into OpenAI’s core business with developers and enterprises. The people who write checks, not the people who make fun clips.
Sam Altman faced a choice: keep funding a $365-million-per-year experiment that users were abandoning, or redirect that compute toward the fight that actually matters. He chose to compete. Sora was collateral damage.
The $33 Million Ghost
Sora isn’t alone.
Three days before Sora’s shutdown hit the news, another AI product died quietly. Yupp.ai, a crowdsourced AI feedback platform, shut down less than a year after launching. They’d raised $33 million from a16z crypto’s Chris Dixon. They had backing from Google DeepMind’s chief scientist Jeff Dean, Twitter co-founder Biz Stone, and Perplexity’s CEO.
1.3 million users had signed up. Millions of preferences were being collected monthly. On paper, the metrics looked alive. But the actual business? The idea was to sell anonymized user preference data back to AI model makers. The problem: by the time Yupp was ready to sell, every major AI lab was building those feedback loops internally. The product’s entire value proposition evaporated before they could monetize it.
$33 million. World-class investors. Real user numbers. Dead in under a year.
And if we zoom out further, the pattern gets louder. Humane raised $230 million for an AI pin that reviewers destroyed and users returned. Rabbit’s R1 device was the star of CES 2024 and a punchline by summer. The entire AI wrapper category collapsed because building on top of someone else’s model turned out to be building on quicksand.
The Pattern Nobody Wants to Admit
There’s a specific disease spreading through AI startups right now. I’d call it Technology Infatuation Syndrome.
It works like this: a team builds something technically impressive. Investors see the demo and lose their minds. Money pours in. Press coverage follows. Everyone assumes that because the technology is amazing, customers will come.
They don’t.
Because here’s the thing about product-market fit that $122 billion apparently can’t teach you: people don’t buy technology. People buy solutions to problems they already have. And “I need to generate a 10-second AI video” is a problem that approximately nobody was waking up at 3am worrying about.
Sora solved a problem that was interesting but not urgent. Not painful. Not something people would pay real money for on a Tuesday afternoon. The technology was a solution looking for a problem, and $1 million a day couldn’t conjure that problem into existence.
There’s a reason the 90-day revenue rule is replacing traditional MVP culture in smart startup circles. The founders who survive aren’t the ones who build the most impressive demos. They’re the ones who find someone willing to pay before writing a single line of code.
What This Actually Means If You’re Building Something
I know what some of you are thinking. “But Liran, I’m not OpenAI. I don’t have $122 billion. This doesn’t apply to me.”
It applies to you more than anyone.
If OpenAI, with effectively unlimited resources, couldn’t sustain a product that people didn’t urgently need, what makes you think you can? You have less money, less time, less margin for error. Which means your product-market fit tolerance is basically zero.
So here’s what I’d take from this.
Impressive demos are warning signs, not green lights. When everyone in the room says “wow” at your demo, that’s the most dangerous moment. Awe is not demand. Excitement is not willingness to pay. The Sora demo videos made people’s jaws drop. Those same people never came back to generate a second video.
If your business model depends on users “eventually” paying, you don’t have a business model. Sora was free for ChatGPT Plus subscribers. Yupp was free for users. In both cases, the plan was to monetize later. Later never came. Later is where startups go to die.
Watch what big companies kill, not what they launch. OpenAI killing Sora tells you something the launch never could: that AI video generation as a standalone consumer product doesn’t have a business. The same way vibe coding tools showed us that easy-to-build means easy-to-replicate, Sora’s death shows us that impressive-to-watch doesn’t mean valuable-to-use.
Urgency beats spectacle. The AI products actually winning right now are boring. Coding assistants. Document processing. Customer support automation. Sales copilots. Nobody makes a viral tweet about their AI invoice processor. But people pay for it. Every month. Without being reminded.
The Uncomfortable Question
I’ll be honest, I’ve changed my mind on this three times while writing it.
Part of me thinks the Sora shutdown is just smart capital allocation. Altman saw the numbers, made a cold business decision, and moved on. Happens every day in tech.
But another part of me wonders if we’re watching something bigger. A market correction in what AI products people actually want versus what founders and investors think they should want. The gap between “this is cool” and “I need this” has never been wider. And companies with billions in the bank are finding out the hard way that you can’t fill that gap with money.
OpenAI will be fine. They still have the most powerful AI lab on the planet, the best-funded position in history, and a CEO who clearly knows when to cut losses.
But the thousands of founders building AI products right now? The ones who saw Sora’s launch and thought “if OpenAI is in video, there must be something there”? They just lost their signal. The lighthouse they were navigating toward went dark.
And maybe that’s the real lesson. Stop following what the giants launch. Start following what they kill.
That’s where the truth is.
