The Keys to the Kingdom

Don't be distracted by the billion dollars.

When Disney announced its OpenAI partnership last week, every headline led with the $1 billion investment. Bloomberg ran the number. CNBC ran the number. Marketing press ran the number. The tech press dutifully reported that Disney was making a significant bet on artificial intelligence.

That's the wrong story. And the wrong number.

Disney's annual revenue is $90 billion. Their market capitalisation hovers around $200 billion. A billion-dollars is the kind of sum Bob Iger finds between the sofa cushions at a board meeting. Disney spends more than that on theme park maintenance.

The real investment is 100 years of irreplaceable intellectual property. Mickey Mouse, who has been printing money since 1928. The Marvel Cinematic Universe, which has generated $30 billion in box office alone. Pixar's entire catalogue of characters that make grown adults weep (don't get me started on Coco..) . Star Wars, the franchise that taught Hollywood what a billion-dollar opening weekend looks like and my personally preferred galaxy to this one.

More than 200 characters, now licensed exclusively to power OpenAI's Sora video platform and ChatGPT Images. The most valuable intellectual property portfolio in entertainment history, handed to a company that didn't exist a decade ago.

That's a hundred-billion-dollar bet. Maybe more. And Disney doesn't make bets like that unless they see something the rest of the market is missing.

To understand why Disney made this move, you need to understand what just happened in the AI industry. And what just happened isn't what anyone predicted three years ago.

For three years, the AI race operated on a simple, seductive assumption: the company with the most compute would win. More GPUs meant bigger models. Bigger models meant higher benchmarks. Higher benchmarks meant market dominance. The logic seemed irrefutable, and it launched an arms race unlike anything technology has seen.

$500 billion committed to Project Stargate. $100 billion in annual infrastructure spending across hyperscalers. NVIDIA's market cap tripling, then tripling again, on insatiable GPU demand. Every AI company fighting for the same scarce resource of compute capacity because conventional wisdom said that more parameters meant better intelligence meant inevitable victory.

The entire industry organised itself around a single bet: that model quality would remain the decisive battleground.

Then Google woke up and caught up.

Gemini 3 now matches or exceeds GPT-5 on every major benchmark. Not by a little. Decisively. The gap that OpenAI spent billions building and defending, the gap that justified their $150 billion valuation, evaporated in a single product cycle. Three weeks ago, Sam Altman declared "Code Red," the company's highest internal threat level. The Information reported executives were told to brace for "rough vibes" and "temporary economic headwinds."

Here's what $600 billion in GPUs actually bought: convergence. When everyone invests hundreds of billions in the same approach, they arrive at similar destinations. The race for the best general-purpose model ended not with a winner, but with a tie.

And in a tie, distribution decides everything.

Google's AI Overviews reaches 2 billion users monthly. Two billion. Gemini has 650 million monthly active users. That's not marketing. That's not user acquisition. That's distribution embedded in Search, Gmail, Android, Chrome, YouTube, and Workspace surfaces billions of people already use before they've made any conscious decision about AI.

When Google adds AI capabilities to Gmail, every Gmail user becomes a Gemini user. No download required. No account creation. No friction whatsoever. It's a software update to services people already depend on.

OpenAI has 800 million weekly active users and superior brand recognition. ChatGPT is practically a verb. But brand recognition without distribution is a slowly emptying gas tank. You can be the most recognised name in the market and still lose to the competitor who owns the surfaces.

The compute race produced a tie. The distribution race was over before it started.

There's a deeper problem facing every foundation model company, and it explains why Disney's intellectual property matters more than their cash.

The models have learned everything.

Every public webpage. Every digitised book. Every Wikipedia article. Every Reddit thread. Every open-source code repository. Every academic paper. Every news archive. The foundational models have consumed the world's available information. They've ingested human knowledge at a scale that would have seemed like science fiction a decade ago.

And they've reached the edges of what public data can teach them.

This is the scarcity nobody saw coming. For years, the constraint was compute, not enough GPUs to train larger models. Now the constraint is shifting. You can have all the GPUs in the world, but if you've already trained on every publicly available dataset, what comes next?

Two frontiers remain for model improvement. Only two.

The first is proprietary intellectual property or content so valuable that its owners have protected it from the training data free-for-all. Disney's characters. Premium journalism behind paywalls. Licensed music catalogues. The creative works that lawyers have successfully kept out of the training sets. This content exists in a protected bubble that foundation models haven't been able to touch.

The second is user-generated content created with that proprietary IP. Every fan video made with Sora using Marvel characters. Every image generated through ChatGPT featuring Pixar worlds. Every creative experiment a user runs with licensed content. Every interaction becomes training signal, and the training signal is exclusive to the platform that enabled it.

This is what Disney understood before most observers. They're not licensing their IP to OpenAI for the revenue share, though the revenue share will be substantial. They're not doing it for the $1 billion investment, though a billion dollars is nice to have. They're licensing it to create a proprietary flywheel that their competitors cannot replicate.

Exclusive content enables exclusive user behaviour. Exclusive user behaviour generates exclusive data. Exclusive data improves the exclusive experience. The improved experience attracts more users, who generate more exclusive data, which further improves the experience.

Google can match OpenAI's model quality. They've already proven that. Google cannot replicate this flywheel. They don't have the content rights to start it spinning.

OpenAI's strategic problem has two parts, and the partnerships emerging over 2025 solve both with surgical precision.

The first problem is distribution. Apple provided the answer. The OpenAI-Apple partnership integrated ChatGPT directly into iOS, iPadOS, and macOS—reaching billions of devices without Google's permission or participation. When you ask Siri a question it can't handle, it routes to ChatGPT. When you use Apple's writing tools, ChatGPT powers the intelligence. The integration is so deep that many users will access OpenAI's technology without knowing they're doing so.

Apple didn't pay OpenAI for this privilege. They didn't need to. Distribution was the compensation. For OpenAI, access to every iPhone on Earth is worth more than any cheque Apple could have written.

The second problem is stickiness. Disney provides the answer. You can get generic AI anywhere. Every platform will summarise your documents, generate your images, answer your questions. That's table stakes now. Commodity capability.

You can only get Darth Vader on Sora. You can only create videos with Marvel characters on OpenAI's platform. You can only play with Pixar's worlds through ChatGPT. The content is exclusive, the experiences are exclusive, and the exclusivity is legally enforced.

The strategic logic is elegant in its simplicity. Apple solves how users reach OpenAI. Disney solves why they stay. Distribution and desire. Access and addiction. The complete package for competing against a company that owns the internet's surfaces.

This is what the Disney deal actually signals: a new front in the AI wars: content ecosystems. Exclusive intellectual property. Proprietary training data. User-generated content that can only be created on your platform. Experiences that justify the friction of choosing one AI over the embedded alternative.

This is the front where Google's advantages matter less. Google has distribution, but they don't have Disney's characters. They have compute efficiency (their TPU chips run 4-6x more cost-effectively than NVIDIA GPUs), but they can't license Star Wars. They have 2 billion users, but those users can't create Marvel videos on Gemini.

On the same day Disney announced the OpenAI partnership, they sent cease-and-desist letters to Google alleging "massive scale" copyright infringement from training models on Disney content without authorisation. This isn't contradiction. This is the playbook executed with precision timing.

License exclusively to your chosen partner. Litigate everyone else into retreat.

Disney picked their horse in the AI race and immediately began kneecapping the alternatives. The partnership isn't just about what OpenAI gains: access to irreplaceable content.

The metaphor writes itself. For a century, Disney has called its theme parks "The Magic Kingdom." The company built an empire on controlling access to beloved characters and stories by licensing selectively, protecting aggressively, monetising relentlessly. Nobody has ever been better at extracting value from intellectual property.

Now they've handed the keys to that kingdom to OpenAI.

Not because OpenAI paid the most. Google has deeper pockets and would have written a larger cheque. Not because Sam Altman is a better negotiator than Sundar Pichai. Both are formidable.

Disney chose OpenAI because they looked at the AI landscape and concluded that OpenAI needed what they had more than Google did. Google is winning. Google has distribution, cost advantages, and model parity. What Google doesn't have is a vulnerability that Disney can exploit.

OpenAI is vulnerable. OpenAI needs differentiation. OpenAI will pay, not just in cash but in equity, in integration, in partnership terms, whatever it takes to build moats that Google cannot cross.

That vulnerability made OpenAI the better partner. Disney's IP is worth more to a company fighting for survival than to a company cruising toward dominance. And Disney structured the deal accordingly.

The billion dollars will be forgotten by next quarter's earnings call. The IP licensing will reshape competition for a decade.

When the compute race produced a tie, the keys to the kingdom became the new currency. Disney had them. OpenAI needed them. Google, for all its advantages in every other dimension of this competition, found itself locked out.

That's the story the headlines missed. Not the billion dollars. The hundred billion in IP. Not the investment. The strategic desperation it reveals. Not what Disney gave. What Google lost.

The kingdom has new gates. And Disney just decided who gets to walk through them.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 15 December 2025 in the Brandflow newsletter on LinkedIn.