You've watched an OpenAI keynote. Sam Altman walks onstage, the crowd holds its breath, a new product appears. The hype machine roars to life. Partnerships get signed. Developers start building. And somewhere, a product manager quietly starts a countdown timer set to six months.

On March 24, 2026, OpenAI killed Sora — its AI video generation tool. For the non-technical: you type a sentence, the AI spits out a video clip. OpenAI launched it in December 2025. Total lifespan: six months. Disney's $1 billion partnership collapsed the same day. And Altman told staff he's stepping away from safety oversight to focus on fundraising and building data centers — the massive server warehouses that power AI.

If any of this surprised you, you haven't been paying attention.

The graveyard has a pattern

Sora isn't OpenAI's first corpse. Let me walk you through the cemetery.

ChatGPT Plugins (2023–2024). Plugins were mini-apps that lived inside ChatGPT — like browser extensions, but for a chatbot. OpenAI launched them with developer fanfare. Altman himself admitted they lacked product-market fit — meaning nobody actually wanted to use them the way OpenAI imagined. Killed in April 2024. Over a thousand developers left holding code that connected to nothing.

GPT Store (2024–?). Custom GPTs let you build your own specialized ChatGPT for a specific task — "an AI that only answers cooking questions," that sort of thing. OpenAI announced this like it would become the App Store of AI. Hundreds of thousands of GPTs appeared. Then silence. Not officially dead, but functionally abandoned. A zombie product shuffling through the interface.

Sora (December 2025 – March 2026). The flagship. Disney signed on. Six months later, both are gone. The official reason? "Focusing on world simulation research." The real reason, per TechCrunch: Sora burned through roughly $15 million a day in inference costs — that's the processing power needed every time someone generates a video. Total lifetime revenue from in-app purchases? $2.1 million. You don't need a spreadsheet to see the problem.

The pattern: announce with maximum hype → attract partners and developers → underinvest → kill it when it doesn't print money within two quarters.

This isn't a technology failure

Sora could generate impressive video. The failure is strategic. OpenAI treats every product like a startup pitch deck: build excitement, secure partnerships, pivot the moment the unit economics — revenue-per-user minus cost-per-user — don't add up.

Except they're not a startup anymore. They're raising capital at a $300B+ valuation. They had Disney making billion-dollar decisions based on their roadmap. Had. Disney reportedly found out about the shutdown less than an hour before the public announcement. That's how you treat a billion-dollar partner — with less notice than a DoorDash delivery update.

The compute excuse is rich. Altman is literally building data centers across the country. OpenAI has more processing power available than virtually anyone except Google. If they can't afford to run Sora, the product's economics were fundamentally broken — they don't lack servers, they lack a business model.

The counterargument, because I'm fair like that

Killing underperforming products is normal business. Google murders products so frequently there's a dedicated memorial website. Meta killed entire hardware divisions. Apple killed the car project after a decade.

Sora faced real problems beyond OpenAI's control. Copyright law for AI-generated video is a legal minefield. Deepfake concerns — AI-generated fake videos of real people — create massive regulatory risk. The market for "type a sentence, get a video" turned out smaller and more price-sensitive than the pitch decks claimed.

Disney pulling out might say more about Disney than OpenAI. They're notoriously conservative with tech partnerships and have walked away from bigger deals for smaller reasons. According to Deadline, no money actually changed hands — they never finalized the deal.

And Altman stepping back from safety? You could read it as "he doesn't care about safety" or as "the CEO of a company scaling to 8,000 employees delegates to his Chief Risk Officer like a normal CEO would." Safety now reports to Mark Chen. That's how org charts work at scale.

What this means for you

The charitable reading: OpenAI is figuring out its identity in real-time, and killing products beats zombie-maintaining them.

The realistic reading: OpenAI has a partner reliability problem.

If you build on OpenAI's platform — whether you're an enterprise, a developer, or a major entertainment conglomerate — you're betting the thing you're building on will exist in six months. That's not a bet serious businesses enjoy making.

Contrast this with Anthropic: fewer products, longer timelines, zero splashy consumer launches. Claude Code has been quietly improving for over a year. Anthropic developed MCP (Model Context Protocol — a universal plug standard that lets AI tools connect to different data sources, like USB but for software) quietly and donated it to a foundation. Boring. And boring is exactly what you want from infrastructure your business depends on.

Google kills products too, but Gemini keeps expanding. Their enterprise API — the connection point where businesses plug into the AI — isn't disappearing tomorrow.

OpenAI's product strategy is "throw spaghetti at the wall and scrape it off after six months." That works when you're pre-revenue and experimenting. When you're the most-valued AI company on Earth with billion-dollar partnerships, it's a liability.

The six-month clock

Sora is dead. The Disney deal is dead. OpenAI's next model is reportedly codenamed "Spud." It'll be announced with maximum fanfare, maximum partnerships, and a six-month clock already ticking in the background.

RIP Sora, December 2025 – March 2026. You died as you lived: burning $15 million a day and confusing everyone about what you were actually for.