When a hyperscaler deploys $100B+ annual capex, Free Cash Flow (OCF minus capex) craters even as the underlying business remains highly cash-generative. FCF-based valuation multiples (EV/FCF, FCF yield) become meaningless or deeply misleading precisely when the investment opportunity is most attractive. Operating Cash Flow — which reflects business cash generation before the capex allocation decision — is the correct anchor. This is distinct from the EPS trough pattern: the EPS/P/E distortion is driven by depreciation timing on the income statement; the FCF distortion is a cash flow statement phenomenon driven entirely by the magnitude of capex spending.
For any hyperscaler or AI infrastructure company in an active capex buildout cycle: (1) set EV/FCF aside entirely — it will systematically mislead; (2) use EV/OCF as the primary cash flow valuation anchor; (3) verify that OCF itself is growing (not just stable) — a growing OCF alongside a collapsing FCF is healthy; a flat or declining OCF alongside a collapsing FCF is a warning. The companion check: if management repeatedly deflects analyst questions about a FCF floor, treat that as a deliberate prioritization signal rather than evasiveness — they're telling you FCF is consciously being subordinated to market share capture. Pair this with the backlog/demand validation check (is capex being pulled by contracted demand, or pushed speculatively?) before accepting the thesis.