ARKHOSReports

Disney’s AI lesson: an equity stake is not control of the product

From the Sora investment that did not proceed to the November 2026 Adobe alliance

On 9 November, Adobe and Disney announced a multi-year alliance around custom AI models and creative workflows. It follows Disney’s May disclosure that its planned investment in OpenAI would not proceed after Sora’s shutdown. Together, these developments raise a more useful question than whether Disney is for or against AI: what exactly should a company own, license and be able to move? [1–2]

First, separate an announcement from a completed investment

The December 2025 plan combined a $1 billion OpenAI equity investment with a three-year licence covering more than 200 characters and access to AI services. It excluded performers’ voices and likenesses. The announcement expressly required definitive agreements, approvals and closing conditions. [3]

Disney subsequently said the investment would not proceed because OpenAI had chosen to shut Sora, while leaving potential commercial opportunities open. OpenAI confirms that Sora’s web/app service ended on 26 April and its API on 24 September 2026. [2,4]

That record does not support calling this a billion-dollar investment loss. Nor does it establish that Disney incurred no preparatory costs. It tells us the planned investment did not proceed. The detailed allocation of rights, termination consequences and costs is not fully disclosed in these sources.

The risk was not only whether AI could generate a compelling video

The proposed commercial logic connected several different assets: Disney’s intellectual property, a provider’s product roadmap, user participation and distribution. A technical demonstration could validate generation quality without validating the durability of that whole chain.

Owning equity in a technology company is different from controlling the future of a particular product. Even a valuable provider can discontinue a service that matters to an individual customer. Nothing in the public announcement establishes that Disney had a product-continuity veto, transferable model access or a guaranteed substitute.

This is not proof that Disney’s board omitted those protections. The contracts are not public. It is the exposure a board should identify before treating a strategic shareholding as protection for an operating dependency.

The November Adobe announcement reveals a different use of the same scarce asset

Adobe says the alliance will tailor models to Disney’s creative properties and support its production tools; Imagineering is already using sketch-to-image and custom-image workflows. A narrower collaboration had been announced in June. These are disclosed activities and intentions, not published evidence of financial return. [1,5]

The contrast is meaningful, but not a like-for-like vendor replacement. Sora’s announced proposition included fans generating licensed-character content. The Adobe work emphasises creative teams using customised capabilities within their workflow. One seeks value partly through external participation and distribution; the other seeks it partly through internal creation and execution.

Our analysis: in the second model, distinctive assets and the approval process can play a larger role in determining what becomes usable output. That could make the technology more complementary to Disney’s existing capabilities. It does not prove legal ownership of every model component, freedom from supplier dependence, or better economics.

“Custom” is not synonymous with “portable.” A model tailored to proprietary material may still rely on another company’s infrastructure, interfaces and licensing. Neither public announcement supplies enough detail to conclude that the new arrangement eliminates the original category of dependency risk.

The board decision changes when the assets are separated

Consider an entertainment, tourism or media group in the Gulf with valuable characters, archives or branded experiences. Three decisions should not collapse into one enthusiastic approval:

Access to a capability: Can the team produce an acceptable output reliably? The approval should test creative quality, review burden, provenance and operational usefulness, not just generation speed.

Permission to use strategic assets: Which material may train or customise a system, for which purposes, and under whose approval? The negotiation should address outputs, model access, reuse by others and the consequences of termination. A press release describing a proprietary model does not answer these legal questions.

An investment in the supplier: Would the equity thesis still make sense if the specific product were withdrawn? If not, the investment case is dependent on a commercial assumption that deserves explicit conditions and reassessment. This is a proposed discipline, not a description of Disney’s undisclosed internal process.

The relevant operating metric also changes. More generated concepts are not automatically more value. A production workflow should measure the cost and time of a concept accepted for its intended next stage, including rework and human review. An external-content initiative must establish a separate link to engagement, monetisation or distribution value.

What can we call a win today?

Disney did not proceed with the planned investment, and it continues to pursue AI opportunities. Adobe has secured an announced strategic relationship. Those are observable developments, not proof of realised investor returns or market-share transfer between AI providers.

The potential winners are organisations able to combine distinctive assets, usable workflows and credible alternatives. The potential losers are those that make irreversible commitments on the assumption that access to a product today guarantees access tomorrow. Whether either description fits a particular company requires contractual and operating evidence.

ARKHOS’s conclusion: the important question is not simply which model is strongest. It is which parts of the value chain remain usable and under your control when the provider changes direction.

If your AI partner withdrew the product tomorrow, would you retain a valuable operating asset, or only an investment thesis that no longer fits?

ARKHOS

Sources

[1] Adobe, 9 November 2026, multi-year Disney alliance.
https://news.adobe.com/news/2026/11/adobe-and-the-walt-disney-company-alliance
[2] Disney, 6 May 2026, shareholder communication/earnings release, SEC Exhibit 99.1. Non-proceeding investment and continued exploration of commercial opportunities.
https://www.sec.gov/Archives/edgar/data/1744489/000174448926000036/fy2026_q2xprxex991.htm
[3] Joint announcement, 11 December 2025, published by OpenAI. Conditional transaction, licensing scope and exclusion of talent likenesses/voices.
https://openai.com/index/disney-sora-agreement/
[4] OpenAI Help Center, Sora discontinuation dates. Accessed 12 November 2026.
https://help.openai.com/en/articles/20001152-what-to-know-about-the-sora-discontinuation
[5] Adobe, 16 June 2026, Imagineering collaboration announcement.
https://news.adobe.com/news/2026/06/adobe-and-disney-imagineering-collaborate

Independent corroboration, not used to expand unverified commercial terms:
Reuters, 9 November 2026, Disney/Adobe custom models.
https://www.reuters.com/business/media-telecom/disney-adobe-co-develop-custom-ai-models-tailored-media-giant-2026-11-09/