Buffett on capital allocation
For forty years Buffett has made the same quiet argument about the CFO function — and the cost of not internalising it compounds
For forty years, Warren Buffett has been making the same quiet argument about the CFO function, and most CFOs still haven't fully internalised it: the single most consequential thing finance does is decide how every retained dollar gets deployed. Three Berkshire letters — 1984, 1987 and 2014 — lay out a discipline most finance functions don't have.
The one dollar test (1984)
Every dollar a company keeps is an implicit claim that the firm has a better use for it than the shareholder does. Buffett's test: retain earnings only when at least one dollar of market value will be created for every dollar retained. The relevant comparator isn't your internal hurdle rate — it's what your owner could earn with the dollar if you sent it back.
The accidental capital allocator (1987)
Most CEOs reach the top through marketing, engineering, or operations — then face capital allocation, a job they've never done. Buffett's number: after ten years, a CEO retaining 10% of net worth annually has deployed more than 60% of all the capital at work in the business. If the finance function doesn't own the rigour of that, the accidental version of the job is the one that happens.
Five paths, not one decision (2014)
Berkshire's five uses of a retained dollar — improve the subsidiaries, bolt-on acquisitions, investee growth, buybacks below intrinsic value, the occasional large acquisition — aren't five strategies. They're a taxonomy, and the discipline is weighing every dollar against all five, every time. Most functions are fluent in one or two and quietly default on the rest.
What this means for a 2026 CFO
Every retention decision — capex, working capital, deleveraging, carrying cash — is a capital allocation decision. The hurdle rate is the wrong comparator. The framework belongs in the CFO function, not as a quarterly artefact. AI has lowered the cost of the visible parts of finance; it has not lowered the cost of judgement.
This is a condensed version. Read the full piece at signal-to-noise.co →