Hold a perpetual on the crowded side and you pay funding at every settlement. Nothing has to happen for that cost to accrue — it is charged for time, not for trading.
Which produces a result that surprises people: a correct thesis, entered early and held, can lose to the same thesis entered late and held briefly. The market did what you expected and the carry ate the difference.
The arithmetic
Funding is small per interval and compounds across a hold. Three settlements a day over a multi-week position is a substantial number of payments, and if you are on the crowded side of a strong trend, funding is elevated for exactly as long as the trend persists.
So the cost is highest precisely when conviction is highest — which is the same unhappy alignment as spreads widening when you need them tight.
Why it changes position construction
A spot position has no carry. A perpetual position has a clock. For a view that may take weeks to resolve, spot is often the cheaper expression even without leverage — and options, with their premium paid once rather than per interval, can be cheaper still.
That is a real decision, and it is invisible if funding is treated as a market indicator rather than as a line item in the round trip.
The one case where it inverts
Being on the uncrowded side, where you are paid to hold. That is the structural attraction of cash-and-carry, and more generally it means a contrarian perpetual position is subsidised while a consensus one is taxed.
Which does not make the contrarian side right — it makes it cheaper to be patient on, and patience is the resource this cost consumes.
The rule
Cost the hold before entering: expected funding over your expected horizon, against the move you are playing for. If the carry is a large fraction of the target, the instrument is wrong even if the view is right.