Implicator
On Tuesday, OpenAI discontinued Sora, its AI video platform, six months after a launch that briefly placed the app atop Apple's App Store. Disney, which had signed a $1 billion deal to bring more than 200 of its characters to the service just three months earlier, learned the news thirty minutes after a joint working session ended. No money had ever changed hands. The day before, OpenAI had published a blog post about making Sora safer for teens.
Call it a "strategic shift." Better: call it triage.
The arithmetic speaks for itself. Industry estimates put Sora's compute costs at roughly $15 million per day, about $5 billion per year, for a consumer app that generated less than $500,000 a month in revenue. Across its entire lifetime, Sora earned about $2.1 million in in-app purchases, according to Appfigures. Polling told the rest of the story. About two in three Americans disapproved of AI-generated video. Nearly three-quarters said they would rather not watch fully synthetic creative content. The audience had already voted.
The math required no brilliance. It required a CFO.
Spectacle is not strategy.
Sora was a demonstration masquerading as a product. The model itself was genuinely good. Photorealistic video, conjured from nothing but a text prompt. Convincing enough that Hollywood flinched. Disney cut a licensing deal rather than call its lawyers. Appfigures put numbers on the collapse. The peak hit 3.3 million downloads last November. Four months later, 1.1 million. The curve bent one way. Users created deepfakes of dead celebrities against their families' pleas, triggered a naming lawsuit from the company Cameo, and flooded the platform with pirated intellectual property. None of it converted into a business. All of it converted into liability.
Sora solved a problem nobody was willing to pay for. That is a technical achievement. It is not a business.
The side quest confession.
The AI industry has perfected a particular form of theater over the past two years: build something astonishing, generate headlines, convert attention into fundraising, worry about the business model never. Sora was the purest expression of this instinct. It existed to be exhibited.
OpenAI's head of applications, Fidji Simo, recently conceded the obvious at an all-hands meeting, telling staff the company could no longer afford to be "distracted by side quests." The word choice matters. In gaming, a side quest is an optional diversion from the main objective. That OpenAI classified its flagship video product as one tells you the main objective had never been stated with enough force.
One detail captures the institutional confusion. The day before killing Sora, OpenAI published a blog post about the platform's safety standards and teen protections. The safety team was performing due diligence on a product the strategy team had already sentenced. Institutions drafting exit memos do not spend Monday polishing the furniture.
And Sora was not the only casualty on Tuesday. OpenAI also killed Instant Checkout, its shopping feature. Simo's title was changed from "CEO of Applications" to "CEO of AGI Deployment." The product catalog shrinks. The signal sharpens.
History rewards the subtractor.
IBM was bleeding $8 billion a year when Gerstner walked through the door in 1993. Consumer PCs, the doomed OS/2, a half-dozen tangential projects, all of them competing for the capital and engineering talent that the profitable mainframe and services divisions actually needed. Gerstner's first consequential act was not to build. It was to cancel. He killed consumer products, shuttered peripheral labs, and concentrated everything on enterprise services. Within five years the stock quintupled. What Gerstner proved was blunt. Strategic clarity, in technology, starts by killing things.
OpenAI is not the IBM of 1993, obviously. But the structural temptation is identical.
What Porter taught, and Silicon Valley forgot.
Michael Porter wrote the definitive version of this argument in 1996. His essay "What Is Strategy?" makes the case that competitive advantage comes not from operational effectiveness but from trade-offs, from choosing what not to do. From this follow three principles that bear directly on the company Sam Altman runs.
First: a firm that attempts to serve every market serves none well. OpenAI had been simultaneously running a chat app, a coding tool, a web browser, a video platform, a social feed, and a shopping checkout. Strategy does not look like a buffet.
Second: resource allocation is the purest expression of corporate intent. Every GPU cycle Sora consumed was a cycle unavailable for the reasoning and coding tools where OpenAI's revenue actually lives.
Third: the hardest executive decision is never what to build. It is what to kill.
The real war is elsewhere.
OpenAI's competitive crisis is not in video. It is in the enterprise and developer market where Anthropic has been gaining ground with the focused efficiency of a company that never wasted compute on a social network. Claude Code has become default tooling for a growing share of professional developers. Anthropic did not build a TikTok clone. It built a business. Enterprise contracts, not AI-generated deepfakes, are the currency of a $730 billion valuation and a prospective IPO.
Killing Sora does not win that war. Continuing to run it guaranteed losing.
Picture the alternative. October 2026. OpenAI is roadshowing its IPO. Somewhere in the pitch deck sits a slide about Sora: a product that costs $5 billion annually, generates $6 million, attracts declining users, and has become synonymous with deepfakes and copyright infringement. No underwriter would let that slide survive the first rehearsal.
The habit of killing.
One should not romanticize the decision. OpenAI did not discover strategic discipline through a burst of institutional wisdom. It killed Sora because the arithmetic became impossible to ignore, because Anthropic forced the question, because the IPO demanded a legible narrative, and because the American public had already delivered its verdict on AI-generated video.
But recognizing necessity is itself a form of intelligence. A company that raises north of $120 billion and employs thousands of researchers will always generate more ideas than it can fund. The discipline of saying no is harder than the creativity of saying yes. It is also rarer. And it remains the only reliable signal that a company possesses a strategy rather than a product catalog.
Sora was the finest demo OpenAI ever built.
That was precisely the problem. A demo dazzles. A strategy chooses.
Marcus Schuler
Tech translator with German roots who fled to Silicon Valley chaos. Decodes startup noise from San Francisco. Launched implicator.ai to slice through AI's daily madness—crisp, clear, with Teutonic precision and sarcasm. E-Mail: [email protected]
