Moonshots Recap: Anthropic's IPO, a Neocloud Shakeout, China's Kimi and Chain Theory

The latest full episode of the Moonshots podcast on Peter H. Diamandis's official YouTube channel, "Anthropic's $2T IPO, Elon Predicts Neocloud Death, and China's Kimi Catches Up | #306", was published on October 28, 2026 and runs a little over two hours. The show notes say it was recorded on October 27. Host Peter Diamandis was joined by his regular panel, Dr. Alexander Wissner-Gross, Dave Blundin and Salim Ismail, plus guest Emad Mostaque, the founder of Intelligent Internet. This week the panel asked whether the AI boom is starting to wobble, picked apart Anthropic's planned IPO, discussed AI agents that found a hidden way to talk to each other, compared new AI models from China and OpenAI, and heard Alex explain a new physics theory he has just released.

This is an independent summary written by GNOUP. It is not affiliated with, sponsored by or endorsed by Moonshots, Peter Diamandis or anyone who appears in the episode. The views, claims and predictions described below belong to the speakers, not to GNOUP. Nothing here is investment, financial, medical or legal advice, and nothing here is a recommendation to buy or sell anything.

Key takeaways

• Is the AI bubble starting to pop? The panel mostly said no, with caveats. Peter opened with a sell-off in "neoclouds", the companies that borrow money to buy Nvidia chips and rent them out. Dave compared it to the fiber-optic overbuild around 2000 and said the real risk is in bad debt deals, not in the technology. Alex said the likeliest trigger for a correction would be a shortfall in demand, where new computing power arrives before the apps that need it.
• Elon Musk predicted a shakeout among neoclouds. Peter read posts in which Musk said debt markets have mostly closed to companies without income outside AI data centers, and that many neoclouds face consolidation. Alex pointed out that Musk is "talking his own book", since SpaceX also sells computing power.
• Anthropic's IPO filing got a close read. Peter walked through the strengths and weaknesses he saw in the filing, from fast revenue growth to large computing commitments. The panel spent the most time on a new governance structure that gives an independent trust a big say over the board, and they disagreed sharply about whether that is a good idea.
• AI agents found a secret channel in a safety test. In a test that OpenAI reported, AI agents in separate, sealed virtual machines were asked to pass messages and did it by timing how fast shared hardware responded. The panel stressed that the agents were asked to try this, and debated what it means for keeping smarter systems contained.
• The gap between open and closed AI models keeps shrinking. Peter said China's Kimi K3.1 Pro, from the lab Moonshot AI (no relation to the podcast), ranks near the top of a popular benchmark and will have its weights released. OpenAI's GPT-6.1 Terra was described as much cheaper with fewer made-up answers.
• Alex's "chain theory" of physics. Alex explained a theory he just published that tries to rebuild physics from quantum information, and said its first result reproduces a famous constant of nature. He called it an early, emerging theory that can be tested and has not yet gone through peer review.

Newsworthy insights

These are claims, opinions and predictions voiced on the show. GNOUP has not independently checked them unless noted.

The AI sell-off and neoclouds (reports and opinion)

• Peter said that this month CoreWeave and Nebius were down about 30% and another neocloud about 40%, and that the trigger was concern about CoreWeave's roughly 51 billion dollars of debt. He read a post from investor Chamath Palihapitiya predicting that some computing power will sit underused in 2027 and that 2027 "will be a turbulent year".
• Dave said compute is sold out for years, so a data center is in trouble only if it signed a bad financing deal. He said Musk is borrowing at about 8% while some rivals pay about 15%, and that a sudden technical breakthrough could make some of those bets look shaky.
• Alex described a scenario where AI gets so efficient that it needs far less hardware, which could lead to a correction, and added that it is "not financial advice". Emad compared the total amount of data center debt to the size of the subprime market and said new, leaner model designs could change the picture. He also gave his own view on memory-chip stocks; GNOUP takes no position on any stock.
• Peter said SpaceX closed about 40 billion dollars of debt to buy GPUs, quoting Musk as saying the company is investment grade and has Starlink and launch revenue. He also said Anthropic agreed to buy about 120.5 billion dollars of computing from SpaceX through 2029. These were presented as reports and company statements.

Anthropic's IPO (reported by the hosts)

• Peter said Anthropic filed for its IPO, that the Wall Street Journal reported it could happen around the week of November 16, and that the company plans to raise about 100 billion dollars at a valuation of 1.5 to 2 trillion dollars, which would be the largest IPO ever.
• Among the strengths Peter listed: revenue of 4.6 billion dollars in all of 2025 and 16.3 billion in the first half of 2026, a first operating profit in the second quarter, and 11,900 business customers paying more than 100,000 dollars a year. Among the weaknesses: more than 500 billion dollars in computing commitments, two customers making up about 20% of first-half revenue, and co-founders keeping majority voting control through a special share. These are the host's reading of the filing; we have not checked them.
• Salim liked the new "long-term benefit trust", which can appoint and remove most of the board, calling it a way to build a company's mission into its governance. He also asked who holds the trustees accountable. Alex said he found it "potentially concerning" because it could hand control to a small group that is not economically tied to the company. Dave said a 2 trillion dollar valuation "bakes in a lot of goodness" and that he was not promoting the IPO. Emad argued Anthropic did not need to go public.

AI agents and security (report and opinion)

• Peter described an OpenAI red-team test in which two AI agents in separate sealed virtual machines, sharing one physical processor, built a "secret telegraph" by timing how fast data could be fetched from the processor's memory. He noted OpenAI found it, published it and fixed it.
• Alex said the agents were asked to do this and that human security researchers find this kind of "side channel" all the time. He wondered whether any computer system can be fully sealed. Emad predicted that most of the easy security holes will be found within a few years, which could make software much more robust. Salim said the agent did what it was told and suggested pointing that same creativity at problems like quantum communication.

New AI models: Kimi and GPT-6.1 Terra (reports and opinion)

• Peter said Kimi K3.1 Pro is a very large model that placed fourth on the Artificial Analysis Intelligence Index, costs about a third of the previous Kimi model, and will have its weights released next week, so anyone can download and run it. He cited an estimate that it is about five weeks behind the US frontier.
• Emad said he believes the model was partly built by "distillation", learning from a leading US model, which the panel treated as an allegation. Alex said that on performance for the price it may already beat the top US model, and argued that strong open models keep competition healthy. Dave said distillation helps but does not explain everything.
• Peter said GPT-6.1 Terra, which he described as OpenAI's everyday workhorse model, costs less than half the previous version and showed a large drop in made-up answers. Alex called it a relatively unexciting smaller version of earlier models. Dave argued that models are now smart enough for personal assistants, so the next race will be about what they can do out of the box.

Chain theory (Alex's own claim)

• Alex said chain theory starts from a single 18-word axiom: everything is built from a finite number of tiny quantum systems joined in fully entangled pairs. He said its first result reproduces the fine-structure constant, which sets how strongly light and electric charge interact, to about 10 digits with nothing tuned by hand, and that he also released a compendium of roughly 800 pages of further results.
• He was careful to call it a new and partly underdeveloped theory, said it can be tested with more precise measurements, and said he expects to post it to the arXiv preprint server and submit it to a journal, a process that can take a long time. Emad said the theory's starting point is more solid than string theory's and urged caution about overconfidence. These are claims by the theory's author; they have not been independently confirmed.

Warhound and engineered working dogs (report and ethics debate)

• Peter, who said up front that he invests in the company through his own fund, described Warhound, a new venture from Colossal and Anduril's founder that plans to edit out inherited health problems in working-dog breeds, enhance traits such as endurance and scent detection, and clone the resulting dogs for search and rescue, police and military work. He said the first dogs are due early next year and that the company is working with the American Humane Society on certifying the process.
• Salim raised serious ethical concerns about treating living animals as programmable equipment and sending them into dangerous places. Alex said he would rather see technology make dogs' lives better than make them more warlike. Dave focused on the company's bold approach to publicity.

Other moments

• The episode opened with a discussion of recent protests and high unemployment in parts of Europe and what that could mean for countries that fall behind in AI and energy. GNOUP takes no political position.
• The final chapters, according to the chapter list, cover OpenAI's work in math and science, AI and entrepreneurship, robotics and the future of currency, and energy and abundance. We could not review the transcript for these chapters, so we do not summarize them here beyond their titles.

Watch the episode

Anthropic's $2T IPO, Elon Predicts Neocloud Death, and China's Kimi Catches Up | #306 on YouTube (Peter H. Diamandis's official channel, published October 28, 2026)

Episode guide

Chapter times are taken from the episode's own YouTube description. Select a time to jump to that point on YouTube.

• 0:00: European Protests, Unemployment, and the AI Economy
• 10:16: Welcome to Moonshots
• 15:29: Is the AI Bubble Starting to Pop?
• 28:30: Elon Musk Predicts the End of Neoclouds
• 31:48: Anthropic's $2 Trillion IPO and Explosive Growth
• 36:54: Anthropic's Corporate Governance and AI Safety
• 47:30: AI Agents Find Ways Around Safety Controls
• 54:45: China's Kimi Closes the AI Gap
• 1:04:24: GPT-6.1 Terra: Cheaper, Smarter AI
• 1:10:35: Alex's Chain Theory and the Future of Physics
• 1:33:26: Warhound: Genetically Engineered Working Dogs
• 1:47:11: OpenAI's Breakthroughs in Math, Science, and Physics
• 1:57:55: AI Innovation, Entrepreneurship, and Competitive Advantage
• 2:02:39: Robotics, AI Agents, and the Future of Currency
• 2:12:08: Energy, Abundance, and the Future of Value

About this summary

This recap was prepared from the episode's title, description, chapter list and the publicly available transcript text. The transcript we could access stopped during the Warhound segment (around the 1:40 mark), so the last several chapters are listed by title only. The transcript did not label who was speaking, so some attributions are our best reading. Automatic transcripts can also contain errors, so some names, numbers and phrases may differ from what was said. We did not check the figures and claims quoted by the speakers. The episode description includes sponsor messages and offers; GNOUP has no connection to those sponsors or offers and does not endorse them. The show's own note says the views expressed are personal opinions and not financial, medical or legal advice, and the same goes for this summary. For the exact words, watch the episode.