On the Podcast Circuit, AI Profit Splits Three Ways

The enterprise case

The enterprise argument starts with a filing. ARK Invest analysts, discussing Anthropic's S-1 on the Brainstorm podcast, said the filing shows high revenue growth and high research and development efficiency. Maggie Fanari added figures from the private-market side on the Master Investor Podcast: Anthropic is growing its revenues in the range of 10 times to 16 times year-over-year, and Anthropic is now cash flow positive, according to Fanari. m:…3761 m:…5053

Fanari framed that performance as part of a concentrated bet. RIT Capital Partners has made more than four times its money on its SpaceX investment, and its direct investments linked to the AI theme represent approximately 10 percent of its portfolio. Fanari also said RIT Capital has generated returns close to 10.7 percent since its inception over 35 years ago, and that its discount to net asset value was roughly 20 percent at the end of September. Those are the firm's own reported figures as presented by its chief executive, and they describe a portfolio built around founder-led companies rather than a broad index. m:…5053

The same episode carried a warning about the financing underneath the buildout. ARK analysts said marginal compute rental rates are currently above 50 billion dollars per gigawatt, that SpaceX is building 10 gigawatts of compute capacity in the coming year, and that debt markets are tightening for non-investment grade companies in the neocloud sector. Enterprise profit, in this telling, is real but it sits on top of very expensive capacity. m:…3761

The consumer case

The consumer argument runs in the opposite direction: profit from transactions, not subscriptions or enterprise contracts. ARK analysts said Meta plans to monetize its Muse AI agent through transaction fees on consumer shopping, and that Meta expects to achieve a 10 times increase in compute efficiency for its AI agents over the next year. The prize, as the analysts described it, is large: the total addressable market for AI-mediated commerce is expected to reach 8 trillion dollars by 2030. m:…3761

That case depends on efficiency arriving on schedule and on shoppers actually completing purchases through an agent. Neither point is proved by the podcast discussion, and the analysts' own framing keeps the consumer opportunity in the future tense: an expected efficiency gain, an expected market size. Founders listening for a business model should hear the gap between a transaction-fee plan and transaction-fee revenue. m:…3761

The hardware and defense case

Palmer Luckey made a third case on Bloomberg: profit from building the machines themselves, in the United States. Luckey said Anduril is partnering with Qualcomm to build a 100 percent United States-manufactured laptop, that Anduril is currently at 90 percent United States and allied supply chain for its computer prototypes, and that the company is already in early prototype production for the new laptop. m:…7659

Luckey also argued that current policy misses the point. He said United States semiconductor subsidies are currently failing to capture the full value chain because the United States lacks domestic packaging and assembly capabilities. He described Anduril's software business, Lattice, as the AI brain for its hardware systems, and Project Meridian as a 120-day project focused on the future of warfare in 2040 to 2050. He added that he expects defense spending to remain high regardless of the outcome of the United States midterm elections. m:…7659

The liquidity counterweight

Chris Donahue offered a counterweight to all three profit stories on the Inside the ICE House podcast. His firm manages 700 billion dollars in liquidity assets, and Donahue said clients prioritize the return of their money, meaning liquidity at par, over the return on their money. Federated Hermes is partnering with BNY Mellon and Goldman Sachs to build a tokenized money market fund structure, in a market where there is currently less than 2 billion dollars in tokenized money market funds out of an 8 trillion dollar market. m:…7846

Donahue also described how his firm is handling AI internally: Federated Hermes is training 50 managers at Carnegie Mellon University to integrate AI from the middle of the organization, and Donahue said he believes AI will increase employment rather than decrease it. The firm closed a 700 million dollar private credit deal in Europe during the first half of 2026, and Donahue said the firm expects high single-digit growth for its money market fund business. This suggests a useful discipline for founders: the people holding the cash are watching AI adoption closely, but they are still paid to protect the principal first. m:…7846