Innovational Intelligence™

Disney and Character.AI: Innovation at Bankruptcy Prices

Philippe Boulanger

Burbank, California, October 2, 2026. Karandeep Anand starts at Disney with a title nobody had held before him in the company’s history: chief technology officer. He reports directly to CEO Josh D’Amaro. Ten days later, Character.AI, the company he had been running only weeks earlier, files for Chapter 11 protection in the Delaware bankruptcy court (Cartoon Brew).

The same day, the startup announces that his new employer has made an offer: $20 million for part of its assets (Variety).

If you are reading this from an executive committee seat, that sequence should make you uncomfortable. It describes, with precision, a way of innovating that has become standard practice in large companies: let others take the risks, then come and collect the talent once the bill gets too heavy for them.

Ten days between a badge and a bankruptcy filing

Character.AI was founded in 2021. Its product lets people chat with AI-generated characters, created by the platform or by its users. The audience came quickly, and the startup’s valuation passed a billion dollars (mac4ever).

The business model never caught up. Court filings show losses of $8 to 9 million a month, driven by the cost of computing power and specialized talent (Boursorama). Between January and August 2026, the startup spent $56.1 million on compute, according to Reuters (Business Today).

The second weight is legal. The platform was challenged over the unauthorized use of protected characters, and over psychological harm to teenagers, including cases of suicide. It settled several of those lawsuits in January (Variety). Government investigations into child safety remain open (TechCrunch).

Disney knows this file well. In September 2025, the group demanded that the platform remove all content tied to its franchises (Le Devoir). A year later, it hires the startup’s former boss and steps forward as a buyer. Les Echos framed the deal as Disney rescuing a controversial startup on the brink of bankruptcy (Les Echos).

The word “rescue” deserves a closer look.

What Disney is actually buying

Read the contract before the press release. The $20 million pays for a perpetual, non-exclusive license on part of Character.AI’s intellectual property, without buying the company (Minted). Disney will be able to use the technology, and so will other buyers.

Disney is the “stalking horse bidder”: its offer sets a floor price, and a rival can outbid it until the sale closes, expected in late November or early December (Deadline).

The heart of the deal lies elsewhere. A court filing spotted by Bloomberg shows that Disney will offer jobs to at least 60 Character.AI employees, the majority of the workforce, for an estimated $60 million in annual compensation. The vast majority of those approached have accepted (TechCrunch).

Do the math. Every year, Disney will spend three times the license price on salaries. What it is buying is engineers who know how to run a consumer chatbot for millions of users, and who learned on a live product what works and what gets you sued.

That knowledge appears in no patent. It was financed by investors who will not get it back, since shareholders are paid last in a bankruptcy.

Google already paid full price for the same team

This story has a precedent, and it is almost comic. Noam Shazeer, co-founder of Character.AI, was a researcher at Google. He left after the company refused to release the chatbot he had built, then founded Character.AI. In 2024, Google signed a licensing deal with the startup that included Noam Shazeer’s return, for about $2.7 billion according to the Wall Street Journal (PYMNTS).

Let’s recap. Google had the idea, the inventor and the technology in house. It said no. It then paid billions to get back what it had let walk out the door. Two years later, Disney licenses part of the same technology for $20 million, less than 1% of Google’s check.

That is what I call innovating late. A five-year-old startup explored a market neither giant dared to explore alone. It absorbed the infrastructure costs and the lawsuits. The giants show up afterwards, each at the moment that suits it.

Why big companies outsource risk

I know this mechanism from the inside. When I was CTO of Neopost, our R&D workforce grew from three hundred to eleven hundred people, mostly in development. The innovation or invention part was limited to a few dozen people at best. To capture new trends, I looked for accelerators and venture funds we could work with. In Silicon Valley, an accelerator like 500 Startups could make an investment decision in a few hours. At our company, I was lucky if a decision came within three months. We ended up taking a stake in a Paris fund to access its deal flow (My book, chapter 11).

This book is for CEOs, managers, and innovators ready to challenge everything and create exponential growth.

Be the first to get the english version!

That speed gap explains most of it. A large company is designed to reduce variance: protect existing revenue and avoid lawsuits. A project like Character.AI combines what a risk committee rejects by reflex, unpredictable compute costs and legal exposure involving minors. Google’s refusal is the clearest demonstration.

I wrote that Microsoft, often criticized for its lack of innovation, compensates through partnerships with and acquisitions of small innovative companies, and that startups are proportionally the most innovative if you look at patent filings (My book). I stand by that. It has a counterpart people mention less often: this strategy works because someone else pays for the failure.

Let’s be fair to the objection. Buying is not necessarily an admission of weakness. Acquiring a ready-made capability rather than rebuilding it over three years can be the most rational decision in the world. The problem starts when acquisition becomes the only channel for innovation, and the organization loses the ability to say yes to its own Noam Shazeer.

The acqui-hire, an acquisition that won’t say its name

The structure Disney chose has a name: the reverse acqui-hire. The buyer hires the key talent and pays the startup a license fee, without buying it. In March 2024, Microsoft hired Mustafa Suleyman and part of Inflection’s team this way, against $650 million in licensing fees. Amazon did the same with Adept’s leadership. The FTC in the United States and the CMA in the United Kingdom examined these deals, to check that the structure does not sidestep merger control (Fortune).

The Disney case adds a variant: the acqui-hire in bankruptcy. The price is set under creditor pressure, facing a startup that loses millions every month. All the bargaining power sits with the buyer.

Financially, the deal is clever. Strategically, it says something worrying about innovation culture in large groups: they know how to buy and how to hire, and they have forgotten how to take a risk.

What your executive committee should take from this

At Google, Larry Page screened acquisition projects with the toothbrush test: is this something you will use once or twice a day, and does it make your life better (My book, chapter 12)? Character.AI probably passed the first criterion for millions of users. The second, for some teenagers, is exactly what the lawsuits are about.

Here is an exercise for your next committee meeting. On a scale of 0 to 5, rate your organization on two points:

Your ability to launch an internal project whose viability is unproven, without waiting for a startup to validate it for you.
The delay between a serious idea and a first funded experiment. Do you count it in days, weeks or quarters?

If you score below 3 on the first point, your innovation strategy runs on a checkbook. It works as long as you hold more cash than your competitors, and as long as a startup agrees to take the risks in your place.

Ask yourself, too, who on your teams is playing the role of Noam Shazeer today. Someone may already have proposed the project you will buy back at full price in three years.

The essentials

Remember three things:

Disney pays $20 million for a non-exclusive license and about $60 million a year to hire most of Character.AI’s employees. It is an acqui-hire, closed at the moment the price was lowest.
Google had the idea and the inventor in house, turned the project down, then paid about $2.7 billion to get the team back. The cost of a refusal always gets paid, just later.
Buying innovation is a legitimate strategy. It becomes dangerous when it replaces the ability to experiment internally.

Innovate or agonize, the choice is up to you.

A corporate event, a seminar, an executive committee meeting, or a management committee meeting?

Philippe’s keynotes on innovation and AI are groundbreaking, you’ve been warned!

Complement a powerful keynote with innovative and impactful workshops.

A keynote inspires and raises awareness; workshops transform!

References

acqui-hire, artificial intelligence, bankruptcy, Chapter 11, Character.AI, Disney, Google, innovation strategy, large companies, Noam Shazeer, open innovation, startup acquisition

Author

Philippe Boulanger

Philippe Boulanger, international speaker on innovation and artificial intelligence, author, advisor, mentor and consultant.

Competent Jerks Lose: Why Teams Pick the Lovable Fool

A Harvard Business Review study of 10,000 work relationships found that people prefer the lovable fool to the competent jerk. Here is what that choice costs your innovation, and how to fix it.

Silicon Transistor: The Hot Oil Test That Stunned Engineers

May 10, 1954, Dayton. Experts say a silicon transistor is years away. Gordon Teal dips an amplifier into hot oil and the music keeps playing. An innovation story about the gap between inventing something and bringing it to market.

Humanoid robots: why humans are still cheaper

Robots can handle 34% of U.S. working hours, yet they are profitable for only 0.3% of tasks. Why the gap is so wide, what Anthropic’s study is really worth, and how to run the numbers in your own company before you buy.

Wearable Brain-Computer Interfaces: 5 to 10 Years, One Condition

Wearable brain-computer interfaces could go mainstream in 5 to 10 years, under one condition most articles will skip: fund the evidence, not the gadget. Here is what leaders should demand before equipping their teams.

French Letter or Capote Anglaise: Who Invented the Condom?

Capote anglaise on one side, French letter on the other: each nation pins the condom on its neighbor. From the Minos myth to Falloppio’s 1564 treatise, here is the documented history of its invention, and what it reveals about how we handle uncomfortable topics.

Disney and Character.AI: Innovation at Bankruptcy Prices