What Altman Said on the Mostly Human Podcast
In his first extended interview since OpenAI announced Sora's discontinuation on March 24, Sam Altman appeared on the Mostly Human podcast with journalist Laurie Segall to explain the decision. The conversation, published on April 2, offered the most detailed account yet of why OpenAI chose to walk away from its flagship video generation product.
Altman framed the shutdown as primarily a resource allocation problem. "It's always about compute," he said. He described a situation where OpenAI had "realized something really important is working or about to work so well that we have to stop a bunch of other projects" in order to redirect capacity.
The "something important," according to Altman, is OpenAI's next generation of coding tools and enterprise AI products. He said the company needed to "concentrate our compute and our product capacity into these next generation of automated researchers and companies." In practical terms, this means Sora's GPU resources are being reallocated to products OpenAI believes have a clearer path to revenue.
The interview was notable for its candor. Rather than positioning the shutdown as a strategic evolution, Altman acknowledged it as a hard tradeoff — one that cost OpenAI a major entertainment partnership and left hundreds of thousands of active users without a tool they had come to rely on.
The Numbers Behind the Shutdown
The financial picture behind Sora's closure is stark. According to multiple reports, the product was burning through roughly $1 million per day in compute costs. Its total lifetime revenue, by contrast, was reportedly just $2.1 million — meaning Sora generated less revenue in its entire existence than it cost to run for three days.
User engagement told a similar story. Monthly app downloads peaked at approximately 3.3 million in November 2025, shortly after the public launch. By February 2026, that number had dropped to about 1.1 million — a 66% decline in three months. Active users peaked around one million and subsequently fell below 500,000.
The per-unit economics were also challenging. Each 10-second video clip reportedly cost approximately $1.30 to generate on OpenAI's infrastructure. For a product offering free-tier access and modest subscription pricing, the math did not favor scale.
The product lasted roughly six months in the public market. By several measures, Sora represents one of the fastest rises and falls of any major AI product launch — a trajectory that raises questions about the sustainability of compute-intensive consumer AI tools priced for mass adoption.
Why Altman Walked Away From Engagement
Perhaps the most unexpected part of Altman's explanation was his concern about the kind of product Sora would need to become in order to survive commercially. He described a dynamic where retaining users would have required building engagement-maximizing features — the kind of design patterns that have drawn regulatory and legal scrutiny at other tech companies.
"You could watch a lot of videos, that would have put a series of incentives on us, and would have led to a bunch of decisions to win that we just didn't want to make," Altman said during the interview.
He specifically revealed that OpenAI considered and rejected a proposal to merge Sora into ChatGPT, where video generation would have been one feature within a broader engagement loop. Altman said he was concerned this could fuel the kind of compulsive usage patterns that have drawn criticism and legal action against social media platforms.
Altman referenced recent jury rulings against Meta and Google regarding addictive design in products used by young people. While he did not draw a direct comparison to Sora, the implication was clear: OpenAI did not want to find itself defending similar engagement-driven design choices.
This represents an unusual case of a major tech company walking away from a product specifically to avoid engagement-driven design patterns — though skeptics may note that the financial case for shutdown was already strong on its own.
What Happens to Sora Now
OpenAI has laid out a two-stage shutdown timeline. The Sora app and web experience will close on April 26, 2026. The API will remain available until September 24, 2026, giving developers using Sora in production a longer runway to migrate. OpenAI recommends that all users export their saved content before the app closes.
The research team that built Sora is not being disbanded. According to Altman, the group is transitioning to work on world simulation models aimed at advancing robotics, under a project internally codenamed "Spud." The connection between video generation and robotics is not as distant as it might appear — both involve training models to understand and predict physical dynamics in visual space.
The most high-profile casualty of the shutdown is a planned $1 billion partnership with Walt Disney Co. Announced in December, the deal would have given Sora users access to more than 200 Disney franchises, including Marvel and Star Wars, as reference material for video generation. The partnership collapsed alongside the product.
Altman said he felt "terrible" informing Disney CEO Josh D'Amaro about the decision. He added that OpenAI is still in discussions with Disney about potential collaboration in other areas, though he did not specify what those might involve.
What This Means for AI Video
Sora's departure is already reshaping the competitive landscape. According to Sensor Tower data, Kling AI saw its global weekly active users increase by 4% to an average of 2.6 million following the shutdown announcement. A Bloomberg analysis published on April 1 noted that Kling, Runway, and Vidu are well-positioned to absorb Sora's displaced user base.
The timing of competitor moves suggests the industry had advance notice — or at least strong suspicions — that Sora was in trouble. Google launched Veo 3.1 Lite on March 31, just days after the shutdown announcement, at less than half the cost of its Veo 3.1 Fast tier. On the same day, Runway announced a $10 million venture fund and Builders program, offering 500,000 free API credits to startups building on its video generation platform.
The broader market signal from Sora's failure is that AI video generation, at current compute costs, requires either massive subsidies from adjacent revenue streams or tight integration with paying professional workflows. Companies like Google and ByteDance can absorb video generation costs through their broader advertising and platform businesses. Runway and Pika, meanwhile, are building dedicated creator and enterprise workflows designed to generate revenue that justifies the compute spend.
A standalone consumer video generation app, priced to compete with free social media tools, appears to be a business model the market has rejected — at least for now. Whether that changes as inference costs continue to fall remains an open question, but Sora's six-month run offers a cautionary data point for anyone building in this space.
Stop scrubbing. Start creating.
Wideframe gives your team an AI agent that searches, organizes, and assembles Premiere Pro sequences from your footage. 7-day free trial.
Frequently asked questions
The Sora app and web experience close on April 26, 2026. The API will remain available until September 24, 2026. OpenAI recommends exporting any saved content before the app closes.
According to Sam Altman, the decision was driven by unsustainable compute costs (reportedly $1M per day), a strategic pivot toward enterprise AI and coding tools, and concerns about needing to build addictive engagement features to make Sora commercially viable.
The planned $1 billion partnership, which would have given Sora users access to Disney, Marvel, and Star Wars content, collapsed when OpenAI decided to discontinue Sora. Altman has said the companies are still in talks about other forms of collaboration.
Kling AI, Runway, Google Veo, and Vidu have all seen increased interest since the Sora shutdown announcement. Kling reported a 4% jump in weekly active users to 2.6 million. Google launched the lower-cost Veo 3.1 Lite days after the announcement.