Character.AI goes bankrupt and Nvidia drops to 34% of China's chip market

Four stories from October 13 that are really one story about money: a consumer AI unicorn ran out of it, Nvidia is losing the Chinese market that used to supply it, Apple is betting $299 at a time that AI belongs on a kitchen counter, and the IMF finally put a number on how much of the AI boom is companies promising to pay each other.

Character.AI files for Chapter 11 and Disney sets the floor at $20 million

Character Technologies filed for Chapter 11 in federal court in Delaware, announced Monday October 12, listing estimated assets and liabilities between $50 million and $100 million, and plans to sell all or most of the company under court supervision while the app keeps running.
Disney is the first bidder at $20 million for a non-exclusive, perpetual license to part of the intellectual property plus certain legal claims, and would lend back former Character.AI staff to keep the service running safely; that $20 million is a floor other buyers can beat, the sale is expected to close in late November or early December, and the company says the platform will be wound down if no buyer wants to keep it alive.

Why it mattersThe numbers around this collapse are the lesson. Google paid $2.7 billion in August 2024 just to license Character.AI's models and bring its founders back; the platform hit 20 million monthly active users by October 2024 and raised over $193 million in total. Two years later the whole thing changes hands for less than 1% of what Google paid for the license. Consumer AI has a cost structure nobody has solved: every hour of engagement is an hour of inference you pay for, and engagement was the only thing Character.AI was great at. Add lawsuits over minors' safety, a Texas privacy investigation, and a ban on under-18 users that cut the audience, and the growth story ran out before the unit economics arrived.

Nvidia's China AI-chip share falls to 34% while Huawei reaches 28%

According to an IDC report seen by Reuters, Nvidia shipped 678,000 AI accelerators in China between January and June 2026, down 44% year on year, leaving it with 34% of the data-centre AI chip market. Huawei shipped 572,000 units, up from more than 300,000 a year earlier, for a 28% share.
Behind them: Alibaba's T-Head at 7%, Cambricon 5%, Baidu's Kunlunxin 4% and Hygon 3%, with nine of the 11 Chinese chipmakers IDC tracks growing shipments. The market shipped 2.008 million accelerators in the half: GPUs fell 26% to 991,000 units while non-GPU accelerators rose nearly 80% to 1.016 million, now the larger category by volume, and four Chinese vendors hold 95% of it.

Why it mattersThis is what export controls look like two years in: not a Chinese industry starved of compute, but a Chinese industry that built its own. Jensen Huang has said Washington's curbs handed Huawei a $50 billion market, and the IDC split backs him up — the segment that grew 80% is precisely the one where Chinese designers hold 95%. The strategic consequence is that there is now a second, self-sufficient AI hardware stack, and it is cheaper.

Apple puts Siri AI on a 7-inch screen for $299

Apple announced the HomeView on October 13: a 7-inch square Liquid Retina display with True Tone and 500 nits, at $299 for both the tabletop version (silver, starlight, rose pink, space gray) and the wall-mounted one (space gray, silver). Pre-orders are open and it ships November 13; the wall version needs a Schneider Home Power Mount, from $59.95 for USB-C, $79.95 AC wired, and $129.95 for PoE in early 2027.
It recognizes up to six people and gives each a personalized Home Screen. Siri AI handles world knowledge and music discovery for everyone, while personal context — Messages, Calendar, cross-app actions, FaceTime, intercom — requires an Apple Intelligence-capable iPhone or iPad. It is not sold in the European Union, because Siri AI isn't offered there.

Why it mattersAmazon and Google have sold screens on speakers for a decade; what Apple is selling is the camera that knows which of the six of you is standing there, and an assistant that only becomes useful if you already own the right iPhone. That is the whole strategy: a $299 object whose value is locked to hardware you bought before. The EU absence is the more interesting detail — a flagship Apple device is now launching with a regulatory hole in it.

The IMF puts a number on circular AI deals: $1.4 trillion

In its Global Financial Stability Report published October 13 at the annual meetings in Bangkok, the IMF reviewed 14 publicly disclosed deals since 2023: hyperscaler-led investors put about $300 billion of equity and $650 billion of debt into financially weaker AI firms, and in return secured $1.4 trillion in future revenue commitments through services or chip orders. About $320 billion of that debt sits on the receiving firms' books, and a handful of companies hold most of both sides.
Assistant director Jason Wu called the structure a 'triple whammy'; the report's term is 'wrong-way risk' — investments depreciate precisely when the revenue streams tied to them are impaired. Surrounding leverage figures: hyperscalers now issue 10% of the world's investment-grade bonds, up from under 1% in 2024; hedge funds' gross equity positions plus margin loans are about 15% of advanced-economy GDP, up from 10%; single stocks are 28% of geared ETFs, up from 16%; and banks holding 22% of global banking assets would fall below 7% core capital under stress, up from 18% a year ago.

Why it mattersThe IMF is not calling a bubble; it is describing a plumbing problem. When the same few balance sheets hold the equity, the debt and the revenue promises, a correction does not stay in one place — it arrives three times at the same door. Kristalina Georgieva said on October 12 that AI investment is what keeps global growth near 3%, and the IMF raised its 2026 forecast to 3.1%. Both things are true at once, which is the uncomfortable part.

The big picture

Put the four together and the subject is not intelligence, it is solvency. A platform with 20 million users and two hours of daily engagement could not pay for the compute those hours consumed, and sells for 1% of what a licensing deal once fetched. Nvidia's best customer base built a cheaper substitute while it was locked out. Apple is selling a $299 physical object instead of a subscription. And the IMF has counted $1.4 trillion of revenue that exists mostly as promises between companies that also own each other's debt. The useful question this week is not which model is smartest, but who is actually paying, with what money, and what happens to the rest of the chain when one of them stops.

Go deeper

Sources & further reading

Topics:AI Chips Apple Startups