OpenAI’s Identity Crisis

When I opened Sora this morning, I was met with a flood of strange and disturbing AI-generated videos. On OpenAI’s video app, I scrolled through fabricated scenes of the Iran war and a barrage of fake Donald Trumps blabbering about Jeffrey Epstein. In my least favorite clip, I watched a man deep-fry an infant. The app lets users create fairly realistic-looking AI-generated clips—including of their own likeness—and then post them on a TikTok-like feed. Not all of them are so unsettling, and for better or worse, Sora has been a steady source of internet virality. Within days of its release, it skyrocketed to the top of the App Store.

Now Sora will soon be dead. Yesterday, OpenAI said that it was shutting down the app and terminating public access to its video-generating technology. The decision was seemingly abrupt: Just a few months ago, Disney announced plans to invest $1 billion in OpenAI as part of a licensing deal to bring its characters to Sora, and earlier this week, workers from both companies were apparently still collaborating. (Disney has since retracted its investment plans.) Even some Sora staffers themselves were reportedly caught off guard by the announcement. Online, people eulogized Sora by posting their favorite videos—such as one featuring a column of spinning penguins and another in which Jesus walks on water to win an Olympic gold medal in swimming.

After OpenAI launched the Sora app, in September, Sam Altman predicted that society was about to undergo a stunning artistic revolution. “Creativity could be about to go through a Cambrian explosion,” he wrote online. But such a revolution never materialized. It’s not that people hate AI slop. In fact, if anything, people seem to have a surprising appetite for it—the latest TikTok trend is raunchy telenovelas starring AI-generated fruit. In response to a request for comment, an OpenAI spokesperson pointed me to a public statement that cites “compute demand” as a key factor in the company’s decision. Generating videos is much more costly than generating text is, and Sora has likely been a real financial drain: In the fall, Forbes estimated that Sora might be costing OpenAI millions of dollars daily, and Bill Peebles, who leads Sora, said that the economics were “completely unsustainable.” (OpenAI declined to comment on Forbes’s estimates at the time.)

The decision to quickly spin up a project and then suddenly pull the plug has become a classic OpenAI move. The company has spent the past few years cycling through new product features and business models with spectacular haste in an attempt to find its way to profitability. OpenAI seems to finally be learning that slop is not a business strategy.

Altman has never had a great plan for how OpenAI will make money. “We have no idea how we may one day generate revenue,” Altman said at a 2019 event. He went on to explain that one day, AI would be smart enough that OpenAI would simply ask the computer how to generate an investment return. “You can laugh,” he told a (rightfully) amused audience. “But it is what I actually believe is going to happen.” After ChatGPT’s success a few years later, investors began pouring money into OpenAI, and Altman has done a tremendous job of marshaling investor funds. The start-up is now worth more than Toyota, Coca-Cola, and Disney combined. But investors like to see returns, and so far, OpenAI hasn’t done much to prove that it is capable of generating enough cash to stay out of the red.

[Read: The MySpace dilemma facing ChatGPT]

That’s not to say it hasn’t been trying: Over the past few years, OpenAI has explored just about every business model conceivable. Last summer, Altman described OpenAI as four separate companies—a consumer-tech business, a massive-scale infrastructure project, an AI-research lab, and an incubator for “new stuff,” including hardware. (OpenAI has a corporate partnership with The Atlantic.)

The trouble with trying to do everything is that sometimes you end up doing nothing well. Sora is the latest casualty in a long string of abrupt reversals, about-faces, and seemingly sloppily implemented projects. Last year, Altman announced a massive joint AI-infrastructure build-out with Oracle and Softbank called Stargate, but the effort stalled, reportedly following poor leadership and coordination. Altman said in 2024 that combining ads and AI would be a “last resort” response—but then, earlier this year, the start-up launched an ads initiative. Last fall, OpenAI debuted a shopping feature, which allowed people to buy products directly inside ChatGPT; yesterday, the company announced that it was killing the feature and pivoting to focus on product discovery instead. In January, the company said that the first of its much-awaited devices was “on track” to launch later this year, but weeks later, court filings revealed that the company is unlikely to debut its new hardware before 2027. OpenAI originally banned NSFW content, and then it announced last year that it would make exceptions for such material, even planning a December rollout for erotica, only to later put erotica indefinitely on hold.

Some amount of change in business plans is only natural for any company, let alone one an industry as fast-moving as AI. But when compared with its peers, OpenAI’s strategy is especially chaotic. The company’s plans are seemingly always provisional: No partnership or product road map feels guaranteed to endure. Earlier this year, Nvidia walked back a commitment to invest up to $100 billion in OpenAI. At the time, The Wall Street Journal reported that Nvidia CEO Jensen Huang had concerns with OpenAI’s “lack of discipline” in its business approach. (When asked about the report, Huang said that it was “nonsense” to suggest he was unhappy with OpenAI.)

OpenAI’s haphazard business strategy has left the company to deal with an identity crisis of its own making. OpenAI is losing ground to Anthropic, its chief rival in the AI race, which has stuck with a targeted approach of selling productivity-enhancing AI tools to other businesses. Anthropic has had great success in its steadfast focus on the enterprise market. Now OpenAI is attempting to copy Anthropic’s playbook. “We cannot miss this moment, because we are distracted by side quests,” Fidji Simo, OpenAI’s applications chief, reportedly told staff in a company-wide meeting earlier this month, explaining that the company needs to nail “productivity on the business front.” To do so, OpenAI is planning to nearly double its head count this year, including by hiring a team of specialists who will help other companies adopt its technology. Even at the product level, OpenAI appears to be copying Anthropic—OpenAI is apparently planning to launch a “superapp” to streamline its product offerings into one app, likely an attempt to compete with Anthropic’s Cowork and Claude Code. “We were spreading our efforts across too many apps,” Simo wrote to employees last week. “That fragmentation has been slowing us down and making it harder to hit the quality bar we want.”

After scrolling through Iran deepfakes and Trump slop on Sora this morning, I navigated to Altman’s account on the platform. I was curious to see what the company’s CEO might have to say about the end of Sora. The last time that Altman appears to have posted on the app was six months ago, back when the app launched. Perhaps that should have been a foreboding sign. I continued watching more clips until a pop-up filled my screen. OpenAI wanted to know how using Sora was affecting my mood. The app offered me a choice between “Thumbs-Up” and “Thumbs-Down.” I hit “Thumbs-Down.”