Disney's $1 Billion OpenAI Deal Never Closed. What Happened Next Reveals More

On the evening of March 23, 2026, teams from Disney and OpenAI were working together on a project tied to Sora, OpenAI's AI video app. Thirty minutes after that meeting, according to Reuters, Disney's team learned that OpenAI was dropping the product altogether. The next day the Sora team posted a short message on X: "We're saying goodbye to Sora."

Disney's new CEO, Josh D'Amaro, was six days into the job. The agreement he inherited, a three-year licence for more than 200 Disney characters and a planned $1 billion equity investment in OpenAI, was effectively over before the investment had closed.

It is tempting to file the Disney OpenAI deal under "AI hype, collapsed." The record says something more useful. The investment never closed, so Disney did not lose $1 billion, and the structure of the agreement is the reason. What Disney did over the following seven months, from a larger buyback to a new technology chief to a bid for Character.AI's intellectual property, says more about its thinking on AI than the deal ever did.

What did the Disney–OpenAI deal actually include?

On December 11, 2025, Disney and OpenAI announced a three-year agreement that made Disney the first major content licensing partner on Sora. Sora could generate short, user-prompted videos using more than 200 characters from Disney, Marvel, Pixar and Star Wars, and ChatGPT Images could draw on the same characters. The agreement did not include talent likenesses or voices, and both companies pledged controls against harmful content and respect for individuals' rights over their own voice and likeness.

A curated selection of fan-made Sora videos would stream on Disney+. Disney would also become a major OpenAI customer, building products on its APIs and rolling ChatGPT out to employees. The headline number was a $1 billion equity investment in OpenAI, plus warrants to buy more. The announcement carried one sentence that mattered later: the transaction was subject to definitive agreements, corporate and board approvals, and customary closing conditions.

Then-CEO Bob Iger told CNBC the deal came with roughly a year of exclusivity, after which Disney could license its characters to other AI companies. On Disney's February earnings call he described 30-second videos with no human faces or voices, using about 250 characters by his count, against the 200-plus in the official announcement. He guessed the Disney+ feature would arrive "sometime in fiscal 2026" but gave no firm date.

The business logic was stated on that call. Iger pointed to the huge growth of short-form and user-generated video on platforms such as YouTube, and said curating Sora's output would jumpstart short-form video on Disney+. He also hoped subscribers could eventually make their own videos inside the app. Read that way, it was a distribution bet as much as a technology bet: give fans a sanctioned way to play with the characters inside Disney's own app, instead of watching them circulate elsewhere.

It was also a risk bet. Critics argued that letting users generate videos of Disney characters could cheapen the franchises. Iger, asked about it, said he did not expect any impact on demand for Disney's own programming.

Why did Disney license its characters while fighting other AI companies?

Because Disney was doing both at once, and the dates show it.

Disney and Universal sued the image generator Midjourney in June 2025, and Warner Bros. Discovery later filed its own suit. In the most recent reporting available, the case was still pretrial. In September 2025, Disney sent Character.AI a cease-and-desist letter, and according to Axios the letter said chatbots built on Disney characters were, in some cases, harmful to children. Character.AI said it had removed them.

Then, on Wednesday, December 10, Disney sent Google a cease-and-desist letter alleging infringement on a "massive scale." Google later removed dozens of AI-generated videos featuring Disney characters, Variety reported. The OpenAI announcement came the next morning.

Iger framed it as participation over resistance. Disney, he told CNBC, would "rather participate" in AI's growth than be disrupted by it.

One reading, and it is only a reading, is that licensing and litigation do different jobs. Enforcement stops unauthorised use. A licence builds an authorised channel where Disney sets the terms: which characters, which formats, no voices, no faces. Disney has never described the lawsuits and the licence as one plan, and nothing public proves they were. But the sequence shows a company trying to define what approved AI use of its characters looks like, with OpenAI meant to be the first paying example.

What happened when OpenAI shut Sora down?

OpenAI announced the shutdown on March 24. The Sora app closed on April 26 and the API followed on September 24. A Disney spokesperson said the company respected OpenAI's decision to exit video generation and shift its priorities elsewhere.

Accounts differ on how Disney found out. Reuters, citing a person familiar with the matter, said Disney's team was blindsided 30 minutes after a working session. Sam Altman later said on a podcast, as reported by Variety, that he personally called both D'Amaro and Iger before the announcement, felt "terrible" about it, and that D'Amaro told him he understood. Both accounts can be true, since one describes a working team and the other chief executives, but the public record does not settle the sequence.

Altman's stated reason was resources: OpenAI wanted to concentrate computing power and product capacity elsewhere. Reuters reported that OpenAI executives had debated Sora's future for some time, and that running the app took significant computing power and left other teams with less. Published estimates of Sora's costs and usage vary widely, so this article does not rely on them. Nothing in the public record shows that Disney's meeting caused the shutdown.

Did Disney lose $1 billion?

No. According to Reuters, citing two people familiar with the matter, the transaction never closed and no money changed hands. That matches the conditions in the original announcement. The investment was a stated intention, not a completed transfer.

What Disney lost was time and a product. Iger had told investors that Sora content was coming to Disney+, and that plan ended with Sora. One person told Reuters at the time that the two sides were discussing other ways to partner or invest. Whether any of that went anywhere has not been publicly reported, so it would be wrong to say every tie to OpenAI was cut.

Law firm Ropes & Gray, in its analysis of the deal, noted reports that the licensing terms were agreed before the equity investment closed. It suggested Disney may have had a contractual right not to proceed once Sora ended. That is a lawyer's inference, because the contract is not public. The practical point is clearer. When a commitment is conditional, a partner's product risk does not automatically become your balance-sheet risk.

What did Disney do with the cash it had set aside?

On Disney's fiscal third-quarter call on August 5, CFO Hugh Johnston said the company's 2026 share repurchases, originally guided at about $7 billion, would now be at least $9 billion. The reason, he said, was largely to use "the cash that had been set aside previously for the OpenAI deal" along with expected proceeds from selling Disney's 50% stake in A+E. A Disney filing put those A+E proceeds at about $1.2 billion and said they would fund additional buybacks.

Two cautions apply. Disney has not disclosed how much it had set aside, so the OpenAI share of the roughly $2 billion increase cannot be calculated. And the target had already moved to "at least $8 billion" by the May call, so the increase came in steps. What can be said is narrower and still notable: capital that may have been headed to OpenAI is going back to shareholders, and the CFO said so on the record.

The scale matters too. Disney guided to about $9 billion of capital spending this fiscal year, mostly for parks and cruise ships, and roughly $24 billion in content spending. Set against those figures, a $1 billion equity commitment was large in headlines but modest among Disney's calls on cash.

Is Disney now building AI inside the company?

Disney has not announced a strategy to replace the OpenAI deal, but its recent moves share a pattern.

D'Amaro's public line has been steady. In May he called AI a meaningful long-term opportunity, to be pursued with human creativity at the centre and respect for Disney's own intellectual property. In August he said Disney was using AI in pre- and post-production, personalisation, advertising and park planning, and that the work would stay "artist-driven."

On the same August call he made a broader argument about control. Asked whether Disney should retreat to licensing its content to others, he said a global base of direct subscribers gives Disney first-party data and a place to build new revenue streams, which matters as new technology cycles such as AI emerge. A licensing-only model, he argued, would give that up. He was talking about streaming, not AI partnerships. But the logic is easy to carry over: Disney wants to own the relationship with fans.

Then came two moves within a month. On September 18, Disney named Karandeep Anand, then Character.AI's chief executive, as its first company-wide chief technology officer, effective October 2. He reports directly to D'Amaro and oversees enterprise technology, infrastructure, data, AI platforms, product and engineering. Disney said a number of Character.AI's technical team were expected to join him.

On October 12, Character.AI filed for Chapter 11 in Delaware and asked the court to approve bidding rules that would make a Disney affiliate the opening, or "stalking horse," bidder: $20 million for a non-exclusive, perpetual licence to certain Character.AI intellectual property and claims, as the company described it, plus transition services from former Character.AI staff Disney has hired. The sale is pending, other bidders can top the offer, and the company expects it to close in late November or early December. Disney's bid is for a licence, not for the company or its app. Character.AI has faced lawsuits from Kentucky and Pennsylvania over its chatbots, and has said it will wind down the platform if no buyer agrees to keep it running.

The history is worth noting. In September 2025, Disney demanded that Character.AI take down its characters. A year later, Disney hired its chief executive. Nothing published shows these steps were planned together, or that the OpenAI outcome caused them.

What was the OpenAI deal really betting on?

Read together, the pieces suggest a different lesson from "the partner failed."

The deal bet that the fastest way to put Disney's characters into AI video was to rent access to a platform someone else built, ran and could change. For a few months that looked efficient. Sora's closure exposed the dependency: Disney held a licence, but OpenAI held the product.

Disney's later actions point the other way, toward owning more of the layer between its characters and its fans: a chief technology officer with an AI-platform background, a push to unify its consumer data, and a small licence to technology and rights that a distressed company can no longer afford to hold. Disney has not described this as its thesis. The CTO announcement frames the hire around technology serving creativity. This is a reading the documents allow, not a stated strategy.

Several things remain open. Disney has announced no replacement for Sora-style fan creation. The Character.AI sale is not complete. The Midjourney case is undecided. D'Amaro has said elements of an expanded Disney+ ecosystem should begin rolling out in spring 2027, which gives investors something to measure.

What can other executives learn from Disney's OpenAI deal?

The lessons are practical, and each one comes from what happened here.

First, an announcement is not a commitment. Disney made the $1 billion conditional, and that condition is why the number stayed a headline instead of becoming a loss.

Second, keep exclusivity short. Iger described it as about a year, which limited how long Disney's characters could be tied to one company's roadmap.

Third, assume a partner's product can disappear. OpenAI ended an app it had launched only six months earlier. An agreement should spell out what happens to the money, the licence and the data if a partner shuts down the product.

Whether Disney's response is a smart pivot or a tidy clean-up will show up in three places: how the Character.AI sale ends, what Disney+ delivers in 2027, and whether Disney licenses its characters to another AI company. Until then, the clearest fact in the record is a small one. Disney spent months negotiating a $1 billion partnership and did not pay for it.