Benjamin Graham on CoreWeave

Security Analysis sets the bar for investment, as against speculation, at “safety of principal and a satisfactory return” reached through thorough analysis (Security Analysis) — and the analysis Graham had in mind runs on years of earnings, not a pipeline of signed contracts. CoreWeave’s own numbers make a real case for reading its backlog as exactly that kind of safety: a $99.4 billion revenue backlog, Q1 2026 revenue up 112% year over year to $2.08 billion, and multi-year commitments from Microsoft, Meta, OpenAI, and Anthropic that read like the “definite prospects” a valuation is allowed to weigh when they’re contractually specific rather than merely hoped for.

The other half of the test is harder to clear. A backlog is a promise of future revenue, not the demonstrated earning power Graham required — average earnings “over a period covering at least a full business cycle,” which a company two years into public reporting doesn’t have (The Intelligent Investor). Customer concentration compounds it: 62% from one name, 77% from the top two, at IPO. Graham’s bond-selection discipline treated concentration like this as a fact that widens the required margin of safety, not one that narrows it, and CoreWeave’s own debt — roughly $7.5 billion coming due by the end of 2026 — is serviced by cash flows concentrated in relationships that could unwind faster than the debt amortizes. Founders selling $2.3 billion of stock since the 2025 lockup expired doesn’t settle the split; it’s one more fact for the arithmetic, not a verdict on it. Graham’s margin of safety is built from what a business has already earned. On CoreWeave, that number is still being written.

Where this lens comes from

Benjamin Graham's framework is set out at the lens page, drawn from Security Analysis, The Intelligent Investor. This page applies it to CoreWeave and nothing else — the company's full record is in the dossier.

The other lenses on CoreWeave