No technology wave has absorbed capital as fast as generative AI. In 2025 roughly half of all venture dollars raised worldwide went to AI companies (~$212bn of ~$425bn on Crunchbase's count, reported January 2026). Two private labs reached valuations near or above $850bn in 2026: OpenAI at $852bn post-money in March and Anthropic at $965bn in May (both reported). The four largest US cloud and platform companies were on course to spend roughly $700–800bn of capital expenditure in calendar 2026, most of it on data centres, chips and power (company guidance and analyst roll-ups, reported October 2026; verify). Much of this money flows in a circle. Chipmakers invest in labs that buy their chips. Clouds invest in labs that rent their capacity. Neoclouds borrow against contracts with customers that are themselves funded by the neoclouds' suppliers. This section explains where AI capital comes from, how mega-rounds and "circular" compute deals are structured, how the public-market AI trade works, how valuations are argued for, what history says about investment booms built on real technology, and how an experienced investor separates the parts of the cycle that are building durable value from the parts that are financing each other.
What the study covers
- 1Overview: why AI capital behaves differently
- 2Where the money comes from
- 3Mega-rounds and how they are structured
- 4Circular deals and vendor financing
- 5The infrastructure bill: capex and how it is financed
- 6The public-market AI trade
- 7The bubble debate
- 8Venture math in AI
- 9Exits
- 10Key numbers
- 11Landmark companies and cases
- 12Pitfalls
- 13How an expert thinks about it
- 14Glossary
Technical exhibits
Worked examples, checklists and case studies included with this study.
AI Capital Cycle Worked Example
Stylized and computed in Python: a reverse valuation of a frontier lab, a vendor-financed chip deal, the revenue a gigawatt of AI capacity must earn, and what a venture fund makes on a late-stage AI cheque
AI Investor Diligence Checklist
Questions to ask of an AI start-up, a lab, a neocloud or a listed "AI beneficiary" before investing: revenue quality, compute commitments, circularity, margins, platform risk, governance