Follow the Money

Bubble Thermometer: 15/100 — cool

15/100 (cool): "bubble" said 0× this week (flat from 0); AI basket market spread unknown; $0B of deals announced this month.

Components: talk 0 · markets 50 · money 0 (weights 40/30/30). Each input links to where it comes from: "bubble" in this week's feed (0, prior week 0) · the AI basket against QQQ over a month · $0B announced in 2026-09. An attention index from this site's own data; it measures temperature, not the future.

Daily readings since 2026-09-16. JSON

Money that goes around

Pairs where cash flows both ways — an investment one way, a purchase commitment back. Computed from the deal list, not curated. Open a pair for every deal and its source.

Circularity score

Of the disclosed funding each lab took from named investors listed here, how much came from companies the lab also pays for compute or chips. Round totals from funds are not broken out, so this is the strategic-money share, not a share of everything ever raised. Every number links to the deals behind it.

Announced per month — Anthropic

Face value as announced; the grey stubs are deals later cancelled or never closed, which stay out of the total. Click a month to list its deals.

Biggest counterparties

Announced and reported amounts at face value; "up to" commitments count in full — that is how the headlines count them, and why every row carries its notes. Every company name is a page.

Private valuations (latest verified)

• Anthropic: $965B (2026-05-28) — source
  Series G: $30B at $380B post-money (2026-02-12). Series H: $65B at $965B post-money (2026-05-28; run-rate revenue above $47B). Confidential draft S-1 submitted on 2026-06-01.

4 deals, with sources

Amounts link to the source they come from; companies link to their pages; types, months and statuses filter the page; the # is the deal's anchor.

Deal Watch — reported in the last 45 days (automatic, unverified)

Headlines from the feed go through a model that extracts who-pays-whom; a report appears here only when two different publishers agree on the parties and the amounts agree within 25%. Every publisher name links to the headline. Nothing in this box is in the verified graph above until a data review confirms it against a primary source.

No two-source reports in the window — the extractor needs an AI provider key.

The bubble reads

• AI's $600B Question — Sequoia Capital (David Cahn), 2024-06-20. Annualised AI infrastructure spend implies about $600B of yearly AI revenue is needed to earn a return; most of it is missing.
• The GenAI Divide: State of AI in Business 2025 — MIT NANDA, 2025-07. Despite $30-40B of enterprise GenAI spend, about 95% of pilots show no measurable P&L impact.
• Bridging a $1.5tr Data Center Financing Gap — Morgan Stanley Research, 2025-08. About $2.9T of global data-center capex through 2028, only about half fundable from hyperscaler cash flow; the rest must come from bonds, private credit and securitisation.
• $2 trillion in new revenue needed to fund AI's scaling trend — Bain & Company, 6th Global Technology Report, 2025-09-23. Meeting projected 2030 compute demand needs about $2T of annual AI revenue; the industry is about $800B short.
• IMF and Bank of England join growing chorus warning of an AI bubble — IMF Global Financial Stability Report / Bank of England FPC, via CNBC, 2025-10-09. Both flagged AI-concentrated equity valuations comparable to the 2000 dot-com peak and the risk of an abrupt correction.
• World Outlook: '1999 meets 1990' — Deutsche Bank Research (Jim Reid), via Investing.com, 2026-06-06. Frames 2026 as 1999-style tech enthusiasm colliding with a 1990-style oil shock, with an AI investment boom propping up US growth.