OpenAI Discontinues Sora, Confirms Enterprise-First Spud Model as Anthropic Closes the Revenue Gap

OpenAI is discontinuing Sora, its AI video generation tool, and redirecting the freed compute toward a new enterprise-focused model codenamed Spud. CFO Sarah Friar confirmed both decisions in an AP interview published April 19, describing Spud as the company’s “smartest model yet” and calling the Sora shutdown “a little heartbreaking, but we’re like, ‘OK, it’s not the main event right now.’”

The timing tells the story. Anthropic announced annualized revenue of $30 billion in early April, surpassing OpenAI’s roughly $25 billion for the first time. Three senior OpenAI executives left the company on a single day last week. And Sora, once OpenAI’s most-hyped consumer product since ChatGPT, was costing approximately $1 million per day to operate while losing users. OpenAI is not tweaking its strategy. It is reversing its founding consumer thesis to survive a revenue race it is currently losing.

The Compute Math That Killed Sora

Sora launched to enormous public attention but never translated that attention into sustainable economics. According to The Wall Street Journal, via 80 Level, the tool was burning roughly $1 million per day in inference costs. Active users peaked at approximately one million before declining to fewer than 500,000 by the time OpenAI announced the shutdown on March 24. Some estimates place the daily compute cost as high as $15 million, though the $1 million figure is the most widely cited.

The Sora web and app versions shut down on April 26, with the API following on September 24. Disney, which had entered a partnership with OpenAI around Sora, found out about the shutdown just an hour before the public announcement, according to the WSJ report. The partnership, once valued at up to $1 billion, collapsed.

For Friar, the decision came down to compute allocation. “We need to make sure that our new model that’s coming has enough compute,” she told the AP. That new model is Spud.

What Spud Signals About OpenAI’s Future

OpenAI describes Spud as offering “stronger reasoning, better understanding of intent and dependencies, better follow-through and more reliable output in production,” according to Friar’s AP interview. The language is deliberately enterprise-oriented: “intent,” “dependencies,” “production” — enterprise procurement vocabulary, not consumer chatbot terminology.

Chief Revenue Officer Denise Dresser, the former Slack CEO who was hired three months ago, laid out the strategic logic in a four-page internal memo first reported by The Verge and CNBC. “Multi-product adoption makes us harder to replace,” Dresser wrote. “We should stop thinking like a company with separate product lines. We should think like a platform company with multiple entry points and one integrated enterprise offering.”

Dresser described early enterprise feedback on Spud as “very positive” and positioned it as the intelligence foundation that would lift all of OpenAI’s products. She explicitly framed the competitive battle with Anthropic as a platform war: “You do not want to be a single-product company in a platform war.”

The memo also took a direct shot at Anthropic’s business model: “Their story is built on fear, restriction, and the idea that a small group of elites should control AI.” This echoes CEO Sam Altman’s February statement that “Anthropic serves an expensive product to rich people.”

The Revenue Numbers Behind the Pivot

The enterprise push has already reshaped OpenAI’s revenue mix. When Friar joined as CFO in 2024, enterprise accounted for about 20% of OpenAI’s revenue. It now represents over 40% and is expected to reach parity with consumer revenue by the end of 2026, according to her AP interview.

OpenAI has more than 900 million weekly ChatGPT users, but approximately 95% of them pay nothing. The company is valued at $852 billion and, by multiple accounts, loses more money than it makes. Projections from SaaStr and other analysts suggest OpenAI could lose $14 billion in 2026, with breakeven not expected until around 2030.

That context makes the enterprise pivot not just strategic but existential. Free consumer users are a compute liability, not an asset, unless they convert to paying customers or train models that serve paying customers.

Anthropic’s Structural Advantage

The competitive pressure driving this shift is specific and quantifiable. Anthropic announced annualized revenue of $30 billion in early April 2026, according to multiple reports. That number surpasses OpenAI’s estimated $24-25 billion annualized revenue.

Friar and Dresser both pushed back on the comparison, telling the AP that Anthropic’s figure is “inflated because it doesn’t account for revenue it must share with cloud computing providers Amazon and Google.” But the trend line has shifted regardless. Luke Emberson, a researcher at nonprofit Epoch AI, told the AP: “They’re likely quite close. Certainly the trends show Anthropic is growing much faster than OpenAI. If that continues, they’re likely to cross soon.”

Anthropic entered the enterprise market first and without the compute drag of a massive free consumer product. It never built a Sora. It never had 900 million unpaying users consuming inference cycles. Dresser’s own memo acknowledged that Anthropic’s “coding focus gave them an early wedge” in enterprise.

The number of Anthropic enterprise customers with annualized spending exceeding $1 million doubled from 500 in February 2026 to over 1,000 by early April, according to TradingKey analysis. Anthropic projects positive cash flow by 2028, compared to OpenAI’s estimated 2030 breakeven.

Three Exits, One Day

The strategic shift is costing OpenAI its people, not just its products. On a single day last week, three senior executives departed: Kevin Weil (former CPO, most recently leading OpenAI for Science), Bill Peebles (head of Sora), and Srinivas Narayanan (enterprise CTO), as reported by The Next Web.

The departures follow a pattern. Of OpenAI’s 11 co-founders, only Sam Altman and Greg Brockman remain. At least 12 senior executives left in 2025 alone. Co-founder and chief scientist Ilya Sutskever, CTO Mira Murati, chief research officer Bob McGrew, and co-founder John Schulman (who went directly to Anthropic) are all gone.

The destinations map the competitive landscape. Tim Brooks, who co-led Sora before Peebles, went to Google DeepMind and then to Meta’s Superintelligence Labs. Shengjia Zhao, a key architect of ChatGPT and GPT-4, became chief scientist at Meta Superintelligence Labs. At least seven additional researchers followed the same path to Meta, according to The Next Web.

Weil’s departure is particularly telling. He moved from CPO to lead OpenAI for Science, an initiative that shipped GPT-Rosalind (a life sciences model) the day before his exit was announced. The science research team is now being “decentralized,” absorbed into other groups. The dedicated initiative no longer exists as an independent unit.

The “Subprime AI” Critique

Not everyone is convinced that an enterprise pivot solves the fundamental economics. Author and AI critic Ed Zitron used the AP interview to frame the broader risk. “It’s what I call the subprime AI crisis,” Zitron told the AP. “People built their lives and they built their businesses on top of these companies that, as they try and save money, will start turning the screws.”

The concern is not hypothetical. Anthropic has already imposed rate limits on heavy users, forcing some to wait hours to use Claude. Both companies have built service tiers that reward premium payers. For startups and developers who built products on top of these APIs, rising prices and tightening access represent real operational risk.

Both OpenAI and Anthropic are expected to pursue IPOs in 2026. OpenAI is valued at $852 billion, Anthropic at approximately $380 billion (with VCs reportedly offering $800 billion in a preemptive round). Neither company is profitable. The race to Wall Street is a race to demonstrate that enterprise revenue can eventually cover the extraordinary cost of training and running frontier models.

The Platform Bet

OpenAI’s wager is that Spud, combined with its existing ChatGPT distribution and new enterprise packaging, can recapture momentum against an Anthropic that has been enterprise-native from the start. Dresser’s memo frames it as a platform play: multiple entry points, integrated enterprise offering, making switching costs high enough that customers stay.

The counter-argument is that OpenAI is arriving late to a market Anthropic defined. Anthropic never had to kill a consumer product to fund an enterprise one. It never had to explain why 95% of its users contribute nothing to revenue. And it never had three senior leaders walk out the door in a single afternoon.

Friar, in the AP interview, invoked the lessons of tech history: “Great companies are very good at, in a reasonable period of time, kind of doing that winnowing down and refocusing and it’s super painful.” Altman echoed the same theme on the Mostly Human podcast earlier this month, calling for sharper focus.

The winnowing is well underway. Sora is gone. OpenAI for Science is dissolved. Three leaders associated with those initiatives have left. What remains is a $852 billion company with 900 million users, projected $14 billion in losses, and a new model called Spud that needs to convince enterprise buyers it is worth building on, before Anthropic’s lead becomes permanent.