Anthropic's filing gave podcasts one set of numbers and four arguments

The filing, as one investor reads it

Jamin Ball, a partner at Altimeter, used a Latent Space deep dive into Anthropic's S-1 filing to argue that the company is accidentally profitable. The core of his case is efficiency: Anthropic's revenue grew 14 times over the same period its compute spend for serving models rose only 2.3 times. He also notes that 80% of Anthropic's revenue comes from models that are six months old or younger. m:…3443

Ball's reading also tries to clear up two misconceptions about the filing. He argues that Anthropic's reported net losses are primarily driven by the appreciation of convertible notes held by strategic investors, an accounting effect rather than an operating loss, and that the company's raised capital and cash position — $141 billion raised and $134 billion in cash, on his reading — leave it well funded. He adds that only 27% of Anthropic's compute is currently allocated to revenue-generating inference, which is his way of saying the company has headroom. m:…3443

The valuation argument

TBPN's hosts, John Coogan and Jordi Hays, took the same filing and turned it into a market question. They put Anthropic's IPO valuation at somewhere between $1.2 trillion and $2 trillion, with the betting market cooling: the PolyMarket probability of an Anthropic IPO by November 24 has dropped to 49%. The episode also notes that tech stocks suffered a $2.5 trillion loss in a single day, the worst since June 5, 2026. m:…8494

The gap between Ball's reading and TBPN's is the story. One side sees a business whose economics are improving fast enough to be profitable almost by accident. The other sees a valuation so large that the market cannot agree on it within a trillion dollars, and a listing that may or may not happen this month. Both can be true at once, which is why the filing is being argued over rather than simply reported. m:…3443 m:…8494

The skeptic's reading

Scott Wilson, chief investment officer of the Washington University endowment, gave 20VC the skeptical version. He predicts a shakeout in the AI industry because many companies lack defensible moats, and argues that current private-market valuations for AI companies are detached from fundamentals. Wilson grew the WashU endowment from just under $8 billion to nearly $20 billion, in part by liquidating 36 or 37 hedge funds and shifting to concentrated direct investments, and his private portfolio has compounded in the low to mid-20s over the last decade, against 10% for public equities. m:…4116

Wilson's argument is not that AI is worthless. It is that the companies without a moat — a durable advantage that competitors cannot copy — will be the ones that fall when the money gets tighter. That is a direct challenge to the optimism in Ball's reading: efficiency gains are real, but they may not be a moat if every lab can achieve them. m:…4116

The mood word

Bloomberg Opinion's Shuli Ren, on the Big Take, supplied the week's mood word: FOGO, fear of getting obliterated, replacing FOMO, fear of missing out. She argues that hedge funds and aggressive retail traders using leverage are the primary sellers in the current AI trade unwind, and that the AI financing boom was crowding out other sectors, raising borrowing costs for industries like real estate and retail. m:…9726

Ren also points to politics: Democratic control of Congress is perceived by markets as a threat to the pace of AI data center construction, and UBS estimates that 30% of planned data center projects in the United States could be delayed by political and regulatory uncertainty. Building a data center with one gigawatt of computing power costs at least $40 billion to $50 billion, so a delay is not a rounding error. The podcasts agree on the numbers. They disagree on what the numbers mean — and that disagreement is the market. m:…9726