OpenAI announced Tuesday that it is shutting down Sora, its AI video generation app and API, just six months after the standalone app's splashy debut. The same day, The Information reported that CEO Sam Altman told staff the company has finished pretraining its next major model, codenamed Spud, and expects a strong result within weeks.
Sora is dead. The Disney deal too.
The company was blunt about it. "We're saying goodbye to Sora," OpenAI told CNN, citing the need to redirect compute toward higher priorities. Video capabilities in ChatGPT are also being shelved. An OpenAI spokesperson said the Sora research team would shift to "world simulation research to advance robotics."
The collateral damage is immediate. Disney is walking away from its December deal to invest $1 billion in OpenAI and license characters like Mickey Mouse for use on Sora. That deal was structured entirely in stock warrants, Bloomberg reported, which means neither side had much skin in the game financially. Disney's statement was diplomatically ruthless: the company said it "respects OpenAI's decision to exit the video generation business." Exit. Not pause, not pivot. Exit.
Sora's trajectory tells a familiar story about compute-hungry consumer experiments at AI companies. The app hit one million downloads within five days of its September launch. But interest cratered fast. By December, downloads had dropped 32% month-over-month according to Appfigures data cited by Engadget, and the decline continued into 2026.
So what's Spud?
Nobody outside OpenAI seems to know much. Altman reportedly told employees the model "can really accelerate the economy," which is either a meaningful signal about capability or the kind of vague CEO-speak that precedes a product demo. No details on architecture, capabilities, or whether Spud will power a new product line or replace existing models.
What's clear is the timing. OpenAI is consolidating. The company's IPO prospectus recently leaked, and killing a compute-intensive consumer product that never found sustainable engagement looks a lot like cleaning house for public market investors. OpenAI plans to merge ChatGPT, its Codex coding tool, and its browser into a single "superapp," according to reporting from the Wall Street Journal. Spud may be the model that powers that consolidation.
The AGI branding push
There's a narrative forming, and it's not subtle. Altman told staff that Fidji Simo's product organization is being renamed to "AGI Deployment." Meanwhile, Nvidia CEO Jensen Huang declared on the Lex Fridman podcast this past weekend that he believes AGI has already been achieved.
Huang's definition of AGI is worth scrutinizing. He defines it as an AI capable of creating a billion-dollar company, even temporarily. By that standard, any viral app built with Claude or GPT that briefly generates massive revenue would qualify. Former Tesla AI chief Andrej Karpathy pushed back, calling the claim premature and noting the models still need significant work. Academic researchers remain skeptical, pointing out that current systems still hallucinate, struggle with novel reasoning, and lack common-sense understanding.
But the label matters commercially. At OpenAI specifically, the term "AGI" carries contractual weight, tied to performance benchmarks and risk clauses in its agreements with Microsoft. Declaring AGI means something different depending on whether you're a researcher, a GPU vendor, or a company trying to IPO.
Altman steps back from safety oversight
Buried in The Information's report is a detail that deserves more attention than Spud: Altman has given up direct control over OpenAI's safety and security teams. Safety now reports to Chief Research Officer Mark Chen. Security moves under President Greg Brockman. Altman's stated reason is focus. He wants to concentrate on raising capital, managing supply chains, and building data centers "at unprecedented scale."
That framing is pragmatic. It also means the CEO of the company most publicly associated with AI safety concerns has personally stepped away from overseeing safety. OpenAI has lost multiple senior safety researchers over the past two years, including co-founder Ilya Sutskever. The company's safety committee was restructured in late 2024 to remove Altman and add independent oversight. Now the day-to-day reporting line for safety runs through the research chief, not the CEO.
Whether this is a maturation of the org chart or a signal about priorities depends on how much you trust the framing. Chen is a capable leader who has been at OpenAI since 2018. Brockman co-founded the company. But the optics of the CEO choosing data center construction over safety oversight, at this particular moment, will not go unnoticed by regulators or critics.
OpenAI has not announced timing for Spud's public release. The Sora app and API shutdown timeline is also pending. Disney's $1 billion investment will not close.
