What the Disney-OpenAI Deal Collapse Reveals About AI IP Licensing
Written By: Ryan Morrison
Listen On: YouTube | Spotify | Apple Podcasts
Get the data behind how high-net-worth investors actually allocate.
Long Angle's annual asset allocation report breaks down how members across $5–$50M wealth tiers structure portfolios across public equity, private markets, real estate, and alternatives.
When Disney announced its $1 billion licensing deal with OpenAI in December 2025, it was framed as the new model for how IP holders would coexist with generative AI. Three months later the deal was dead. OpenAI shut down its Sora video product, Disney exited the partnership, and no money ever changed hands.
Michael Sapherstein has spent his career inside the rooms where these contracts get written. He was the first dedicated digital media lawyer at Marvel, joined just before the iPad changed everything about digital comics, and helped package Marvel for its sale to Disney. The conversation surfaces a framework for understanding why the Disney-OpenAI deal was structurally exposed from the start, and what that means for anyone underwriting an IP-rich business.
TL;DR
Disney's $1 billion Sora licensing deal with OpenAI was announced in December 2025 and unwound by March 2026 when OpenAI shut down Sora
No money changed hands before the partnership collapsed, exposing how fragile high-profile AI licensing arrangements can be
Saperstein's framework: every IP licensing deal lives or dies on the rights-reservation language inside the contract, and AI is forcing IP holders to renegotiate assumptions that have held since "all media now known or hereafter invented" became standard
The structural question for investors is not whether AI-IP deals will happen, but how to underwrite businesses whose value depends on rights frameworks still being written
For knowledge-work-heavy businesses, Sapherstein argues AI displaces more than it augments, and that's a portfolio underwriting input, not a future scenario
Table of Contents
The Rights-Reservation Principle That Explains the Fragility
Why AI Companies Are Licensing Instead of Relying on Fair Use
Frequently Asked Questions
Resources Mentioned
Conclusion
What the Disney-OpenAI Deal Actually Was
Disney signed a three-year licensing agreement with OpenAI in December 2025 to bring more than 200 of its characters to Sora alongside a $1 billion equity investment.
The terms were unusually structured. According to Disney's December announcement, OpenAI's Sora video platform would generate user-prompted social videos drawing from a defined set of animated, masked, and creature characters across Disney, Marvel, Pixar, and Star Wars, including costumes, props, vehicles, and environments. ChatGPT Images would have similar access for stills. The agreement explicitly excluded talent likenesses and voices.
The commercial structure went well beyond a standard licensing deal. Disney committed to becoming a major OpenAI customer, deploying APIs across products and ChatGPT for employees. The $1 billion equity investment came with warrants for additional shares. A joint steering committee was formed to monitor user-generated content against a brand appendix defining what Disney would and would not allow.
This was the first time a major Hollywood IP holder had built this kind of multi-layered relationship with a generative AI company. It was meant to be a template. Three months later it was a cautionary tale.
The economics that govern who actually captures value when distribution shifts is the right lens for understanding why this deal was structured the way it was. Disney was hedging across licensing revenue, equity upside, and enterprise customer status simultaneously.
Why the Deal Collapsed in Three Months
OpenAI shut down Sora in March 2026, citing compute costs and weak consumer adoption. Disney exited the partnership before any money changed hands.
Sora launched to a strong start. It hit number one on the iOS App Store and accumulated over a million downloads in its first week. Sustained usage never materialized. By early 2026, active users had fallen below 500,000, and the economics of running a compute-intensive consumer video product no longer justified the investment ahead of OpenAI's expected IPO.
Variety reported Disney exited the partnership in late March, with Hollywood Reporter sources confirming that no funds had been transferred. Disney's official statement was diplomatic, saying the company respected OpenAI's decision to exit the video generation business and shift its priorities elsewhere.
The specific signal in the collapse is more interesting than the news of it. A flagship AI-IP arrangement built on a specific product surface fell apart when the product surface itself disappeared. The contract architecture didn't fail. The product foundation underneath the contract collapsed. That distinction matters for anyone trying to assess the durability of the next round of these deals.
The signal beyond the news cycle.
Beyond Wealth is Long Angle's weekly newsletter covering the decisions high-net-worth investors are actually navigating, from private markets to portfolio construction.
The Rights-Reservation Principle That Explains the Fragility
Every IP licensing deal lives or dies on the rights-reservation language inside the contract. The Disney-OpenAI deal was built on a product surface that was always going to evolve.
Sapherstein's operating framework comes from a discipline most investors never see: how the most valuable contracts inside a media company are written. "If your license doesn't cover all media now known or hereafter invented or the like, you could have a blind spot of technology that may become ubiquitous in the future." That clause exists in modern licensing contracts because of contracts written before it that didn't include it.
The Marvel side of the Spider-Man-Sony arrangement is the canonical example. The original deal predates streaming, predates digital comics, predates mobile gaming. Every renegotiation since has been about reaching backward to reserve rights that should have been reserved the first time. Sapherstein worked on it directly during his Marvel tenure.
The Marvel-iPad inflection is the same story in faster motion. When Sapherstein joined Marvel in 2008, the company hadn't yet built out its digital licensing infrastructure. He had to sign an NDA with Apple to discuss what was coming. The iPad released, Marvel comics rendered in full color and high definition, and the digital comics business changed overnight. Some pre-2010 licenses didn't reserve digital rights cleanly. Licensees ended up with windfalls Marvel never intended to grant.
The implication for AI-IP deals is direct. A licensing arrangement is only as durable as the contract architecture beneath it, and contract architecture has to anticipate technology shifts that haven't happened yet. The Disney-OpenAI deal had clean reservation language. What it didn't have was a product foundation built to last three years.
Why AI Companies Are Licensing Instead of Relying on Fair Use
AI companies are licensing IP because they don't trust their fair use defenses to hold up in court, not because licensing is structurally cheaper.
This is the inversion of the public framing. Mainstream coverage treats the wave of AI licensing deals (News Corp, AP, Universal-Udio, Reddit-Google, Wikipedia, and others) as evidence that the AI industry has matured into responsible partnership. The actual driver is more defensive than that. Professor Jonathan Barnett observed that the fact some model developers are entering content licensing deals suggests they are not as confident in their ability to prevail in infringement litigation. If they were confident, they would not be paying.
Bloomberg Law's docket analysis shows the same pattern from the other side. Media organizations are bifurcating into two camps: those licensing to AI companies and those suing them. The split signals genuine uncertainty about how the courts will rule on training data fair use, and that uncertainty is what makes both sides willing to negotiate.
Sapherstein's view as a former IP licensing attorney is direct: "It just feels unfair that because you built an AI model and you ingested all of this IP that you didn't own that somehow then it's free reign for anybody to do anything they want." That frustration is shaping how the largest IP holders are positioning. The same week Disney announced its OpenAI deal, the company sent a cease-and-desist letter to Google alleging large-scale infringement. Disney has also joined Universal in suing Midjourney and the Chinese AI firm MiniMax for copyright infringement.
What this looks like inside an existing IP-rich asset class is visible in how a $1B music catalog operator is positioning around AI licensing. Music IP is the closest direct analog to character IP. Both are 100-year assets. Both depend on contract architecture that anticipates new distribution. Both are being aggressively repositioned right now.
The underwriting question behind this.
If rights-reservation language is an underwriting input, where do you actually pressure-test it on a specific deal? Long Angle's investments team produces full diligence packages on private offerings, and members challenge the underwriting publicly in a vendor-free forum.
What the Collapse Means for IP-Rich Business Valuations
Investors underwriting IP-rich businesses must evaluate the rights-reservation language behind the revenue, not just the headline licensing announcements.
This is where the conversation gets concrete for anyone holding equity in or evaluating a media or entertainment-adjacent business. Disney's stock has been roughly flat for a decade despite the Marvel-era growth that drove its previous run. Sapherstein doesn't have a clean explanation for why the market refuses to revalue Disney upward, but his observation is that streaming gets too much credit for the company's revenue mix. Consumer products licensing remains a meaningful contributor to Disney's earnings, particularly the toy deals with Lego and Hasbro that operate as near-pure-margin licensing income.
For an investor, the implication is that a media business often looks more like a portfolio of licensing contracts than a streaming company. The valuation should reflect the durability and scope of those contracts, not just the streaming subscriber count. When you're underwriting a company like this, the question is not "what is the streaming revenue?" It's "what does the rights-reservation language in their top 50 licensing contracts actually grant?"
The pro sports rights story is the counterexample of contract architecture done well. MLB Advanced Media was formed because 30 club owners voted to centralize their interactive media rights in a single entity. Sapherstein joined as the second in-house lawyer there, working out of a loft in Chelsea Market when streaming was barely a concept. That centralization decision is what made the streaming infrastructure possible at scale, which in turn made multi-billion-dollar team valuations possible. The rights structure preceded the valuation. Most investors evaluating pro sports today are looking at the wrong end of that causal chain.
The same investor lens applied to software businesses produces a different but related question. The shift from SaaS to AI-native businesses is forcing investors to re-underwrite the durability of revenue streams against a category-level technology shift. The IP-licensing parallel is exact: contract architecture that anticipated one distribution model may not survive the next one.
The Knowledge-Work Argument That Compounds the Risk
Sapherstein argues AI is more likely to displace knowledge work than augment it, which compounds the risk for any portfolio holding professional services or content production businesses.
His view runs against the consensus framing. "The super smart AI folks who are somewhat worried about the future of work, I think have it right more than those who are saying it's just gonna augment your job. You just have to know how to employ it and then your job's fine. I think that's a little naive." From someone who spent two decades inside the legal profession, this is a substantive position, not a hot take.
The bot-to-bot deal negotiation scenario is the operational expression of that view. Sapherstein walks through it directly: AI tools with negotiating models built in, company standard positions and fallback positions encoded, two AIs negotiating with each other in milliseconds and landing on terms both sides' models accept. Lawyers don't draft. They review what the bots produced.
What makes this argument credible coming from him is the lawyer-mindset observation that compounds it. Big law partners earn $3 million a year for decades but rarely show up in active wealth-stage peer communities. Sapherstein's read is that lawyers as a class are trained to be risk-averse, which translates into deference to professional advisors and disengagement from the questions Long Angle members typically gravitate toward. The same mindset that picks a 30-year law firm career picks the target-date 401(k) and stops there.
That's a behavioral observation, but it has portfolio implications. If an entire professional class has structurally underweighted alternatives and active investing while their core revenue-generating activity is the most exposed to AI displacement, the timing of the displacement matters more, not less. The same pattern shows up when you ask whether AI will replace CPAs and other complex professional services, which is a real conversation happening across high-net-worth peer groups right now. Norton Rose Fulbright's analysis of the 2026 AI copyright case trajectory points toward formal licensing as the emerging industry response to fair use uncertainty, which is the same pressure compressed onto the legal profession itself.
Frequently Asked Questions
What was the Disney-OpenAI deal?
Disney and OpenAI signed a three-year licensing agreement in December 2025 that would have allowed OpenAI's Sora video platform to generate user-prompted videos using more than 200 Disney, Marvel, Pixar, and Star Wars characters. The deal also included a $1 billion Disney equity investment in OpenAI and a customer relationship in which Disney would deploy OpenAI's APIs across its products and ChatGPT for employees. It was the first time a major Hollywood IP holder had built this kind of multi-layered relationship with a generative AI company.
Why did Disney cancel its OpenAI deal?
Disney exited the partnership in March 2026 after OpenAI announced it was shutting down Sora, the consumer video product the licensing deal was built around. OpenAI cited extreme compute costs and weak sustained adoption. Active users had dropped below 500,000 by early 2026, and the company is shifting resources toward enterprise products ahead of an expected IPO. Without Sora as a viable product surface, the rationale for the Disney arrangement collapsed.
Did Disney lose money on the OpenAI investment?
No money changed hands before the deal was canceled. The $1 billion equity investment was announced but never finalized, and the licensing agreement had not yet generated any revenue or expense. Disney exited the partnership at no direct financial cost. The reputational and strategic implications are a different question, but the financial loss on the deal itself was effectively zero.
What does the Disney-OpenAI collapse mean for AI content licensing?
The collapse signals that AI-IP licensing arrangements built on specific product surfaces are structurally fragile, even when the contract architecture itself is sound. Other AI licensing deals (with News Corp, Universal, Wikipedia, and others) are continuing, but the Disney-OpenAI outcome shows that flagship deals can unwind quickly when the underlying product strategy changes. Investors and IP holders should evaluate the durability of the product foundation alongside the contract terms.
Will AI replace lawyers and other knowledge workers?
The honest answer is that displacement is more likely than the consensus "augmentation" framing suggests. Sapherstein's view, drawn from two decades inside the legal profession, is that AI tools will eventually handle the lion's share of work currently done by entry-level associates and substantial portions of in-house legal work. The same pressure applies to other knowledge-work-heavy professions. Whether displacement happens in two years or ten matters for portfolio construction, but the direction is clearer than most professional class members are willing to acknowledge.
Final Thoughts
The Disney-OpenAI collapse is being covered as a news story. It deserves to be read as data. The headline AI-IP deals are reshaping the public conversation about how AI companies and content owners coexist, but the durable underwriting question sits one layer below the headlines. It lives in the rights-reservation language inside the contracts, the product foundations the contracts depend on, and the legal framework around fair use that's still being decided in court.
For investors holding IP-rich businesses, that means contract architecture has become a portfolio variable. For IP holders, it means the next wave of AI licensing deals will need to anticipate product churn that the first wave didn't. And for anyone evaluating professional services or content production businesses, the displacement question that Sapherstein names directly is no longer hypothetical. It's an underwriting input.
What the collapse really shows is that the people drafting these contracts are working out a framework in real time. The early deals are going to be messy. The investors who get this right will be the ones who treat rights-reservation language as a portfolio variable, not a footnote.
The peer room behind questions like this one.
The harder underwriting questions don't get answered alone. Long Angle is a vetted community of high-net-worth founders, executives, and investors who pressure-test these decisions in a solicitation-free environment. Members compare notes on private deals, contract architecture, and the underwriting questions that don't have clean public answers. When the AI-IP licensing landscape is still being written, the people you compare notes with matter more than the analysis itself.
Resources Mentioned
Disney's December announcement - Original deal terms
Never miss an episode. New conversations every Wednesday with experienced investors, founders and operators navigating complex wealth decisions. Subscribe for updates.
