Brian Reed: Hey. You see this filing come through?
Eliza Ward: Yeah, just now. Character Technologies — Chapter 11, Delaware bankruptcy court, October 13th. It's real.
Brian Reed: Character.ai. Like — the companion chatbot. Bankruptcy.
Eliza Ward: Voluntary filing. They posted it on their own blog. And the language in that post — the platform will be wound down if no suitable bid emerges. That's not boilerplate. That's them saying this could actually just... stop.
Brian Reed: Wait — a bid. So there's already a buyer on the table?
Eliza Ward: Walt Disney Company. Stalking horse at $20 million for a non-exclusive perpetual license to some of the IP. Not the whole company — the IP.
Brian Reed: Twenty million. For a company that — hang on, wasn't this thing valued in the billions not that long ago?
Eliza Ward: CEO Colin Adams's affidavit has them losing $8 to $9 million a month as of August. That's the number that kind of reframes everything.
Brian Reed: And that $8–9 million monthly loss — that's *after* years of trying to close the gap. Because the affidavit also shows compute at $13.8 million a month in 2024. Against $2.8 million in subscription revenue. Every single month.
Eliza Ward: That's the part that doesn't read like a startup stumble. It reads like — okay, think of a restaurant charging $5 for a meal that costs $25 to cook. More customers doesn't help you. It accelerates the loss.
Brian Reed: And they had 1.7 million app installs in the first week in May 2023. The demand was absolutely real.
Eliza Ward: Which is the whole thing — scale made it worse, not better. Every new user was another meal at a loss.
Brian Reed: The part 404 Media keeps framing — that this proves 'the economics of AI don't work' — but is that what the documents actually show, or is that... reaching a bit?
Eliza Ward: Narrower than that. The documents show *this* cost structure didn't work. $56.1 million on compute between January and August 2026 alone — that's one company's number. Whether it generalizes to the whole sector, I mean, the affidavit isn't making that claim. Colin Adams is describing Character.ai's books, not the industry.
Brian Reed: Right — but they did cut compute from $13.8 million down to $10 million month-on-month by 2025. That's not nothing. And the gap still didn't close.
Eliza Ward: No. And wait — that's actually the more damning detail. They optimized. The revenue line stayed at $2.8 million. The math didn't move.
Brian Reed: So the question isn't why they ran out of money. It's why anyone — Google, investors — thought the subscription model would ever catch the compute bill.
Eliza Ward: That's the take I want to push on — because the circulating version is 'Google bailed them out.' Two-point-seven billion dollars in August 2024. And that just... that reads wrong to me.
Brian Reed: How does it read wrong? That number is real.
Eliza Ward: The number's real. What left with it — that's the part. Noam Shazeer and Daniel De Freitas went back to Google as part of that deal. So the $2.7 billion and the two founders who built the core research, all out the door at the same time.
Brian Reed: And what was left was — Dominic Perella, the general counsel, serving as interim CEO. Then Karandeep Singh Anand. Who then joined Disney the month before the filing.
Eliza Ward: Right. That's three leadership transitions in under two years.
Brian Reed: So the part I actually can't answer — and I don't think we can answer it — is how much of that $2.7 billion licensing value reached Character.ai's operating budget versus flowing to the founders and early investors. Because if it mostly went out as a liquidity event and not into the company's runway, then 'Google rescued them' is just... factually not what happened.
Eliza Ward: We don't have that breakdown. The internal financials aren't public. But the outcome is — 24 months after the Google deal closes, the company files Chapter 11. That trajectory is what we have.
Brian Reed: And OpenAI and Meta didn't sit still during those 24 months. The user base Character.ai needed to monetize — that's exactly who those companies were targeting. So even if the founders had stayed, the competitive pressure was compressing the ceiling.
Eliza Ward: The $20 million Disney bid and what the teen safety settlements tell us about what this IP is actually worth — and who absorbs the loss if the platform just stops — it gets harder there. We'll get there.
Brian Reed: Twenty million is the number that actually stops me cold. Because — okay, the stalking horse bid isn't for the company. It's a non-exclusive perpetual license to certain IP. Disney doesn't own the platform. Doesn't have to run it. Doesn't have to worry about the users.
Eliza Ward: Non-exclusive. Which means other bidders can still come in.
Brian Reed: Right — but the floor is twenty million. For IP that someone once implied was worth a billion-plus. And Karandeep Singh Anand already moved to Disney the month before the filing. With other former employees. Disney confirmed they'll provide transition services. So I mean, the talent and the access to the IP — that left before the bankruptcy was even announced.
Eliza Ward: The structure of that actually matters. Disney didn't acquire Character.ai. They positioned to receive specific rights — and they absorbed the people who understand what those rights are worth — before the court even opened the sale.
Brian Reed: Which — wait, that's not vulture behavior, exactly. That's more like... they already know what they're buying and they set the minimum to keep anyone else from getting a bargain. The sale closes late November, early December, Delaware court. That's soon.
Eliza Ward: What cuts through the IP valuation question is the teen safety timeline. Character.ai barred under-18 users starting November 25th, 2025. Then, in early 2026, they and Google settled lawsuits from families whose kids died by suicide. Colin Adams's affidavit explicitly names 'significant legal, regulatory and operational headwinds.' That's not background detail. That's the CEO telling the court: we cut off our most engaged cohort right when we needed revenue to scale.
Brian Reed: And the practical consequence of that — picture a kid who'd been on the platform for eight months before November 25th, 2025. Conversation histories, characters they'd built real attachment to. If no bid clears the floor and the platform winds down, those records just... what? Deleted? Absorbed into whatever Disney licenses?
Eliza Ward: We don't know. The filing doesn't address user data disposition. That's — actually, that's the unanswered part the court process has to resolve. What the blog post says is: the platform continues operating during the sale. After? Conditional on a bid clearing.
Brian Reed: So the concrete thing to watch before December is whether any bid above twenty million comes in — and whether whoever wins is buying a live platform or just rights to call a Disney character a chatbot.
Eliza Ward: Millions of engaged users. Real engagement. And still, the compute-to-revenue math never moved. So which other consumer AI product is actually running unit economics that close that gap? Because Character.ai had the Google deal, had $2.7 billion in licensing, had two years of runway after that, and it still filed in Delaware on October 13th.
Brian Reed: And that's — yeah, that's the question I don't think either of us can answer. Because we don't have the internal books for OpenAI's consumer products or Meta's. We just have Character.ai's affidavit. But the fact that this is the first public data point — $20 million stalking horse against a multi-billion valuation — that's now the reference number anyone else prices against in a distressed sale.
Eliza Ward: That's fair. That's — actually, I want to be careful there. The Disney bid is non-exclusive IP rights, not the company. So it's not a clean comp for what a whole consumer AI platform is worth. It might just be what Disney thinks certain character rights are worth to Disney.
Brian Reed: Right — but it's still the floor. Delaware court, late November or early December. If no competing bid clears twenty million, that's what the market said. And the blog post is explicit: no suitable bid, the platform winds down. That's not my inference — that's the filing.
Eliza Ward: We leave it there, then. The sale closes by December. Either something clears the floor and the platform survives in some form, or it doesn't. That's the only question the evidence actually supports right now.
