Critical Power (CP)

Critical Power is the highest power a cyclist can sustain in a metabolic steady state — the asymptote the power-duration curve flattens toward as effort duration grows. At or below CP the effort is theoretically sustainable for a long time; above it you're on a countdown, draining your finite work capacity (W'). It's the physiologically-grounded cousin of FTP, derived from a mathematical model of your power-duration curve rather than a single all-out test.

P(t) = W' / t + CP
    (hyperbolic power-duration model; Monod & Scherrer 1965)
    equivalently: t_to_exhaustion = W' / (P − CP)

Monod and Scherrer introduced the critical-power concept in 1965 for isolated muscle groups; later work extended it to whole-body cycling, and modern researchers — notably Jones, Vanhatalo, and Poole — formalized the two-parameter model now in common use. The model has exactly two parameters: CP, the aerobic asymptote in watts, and W', the finite work capacity available above it in kilojoules. Both are estimated by having a rider perform two to five maximal efforts of different durations (classically in the ~2–15 minute range) and fitting them to the hyperbola.

Physiologically, CP marks the boundary between the heavy and severe exercise-intensity domains: it is the highest intensity at which oxygen uptake and blood lactate can still stabilize. Ride above CP and VO2 drives toward maximum while lactate accumulates relentlessly until you're forced to stop. That makes CP a more principled threshold than FTP's operational 'power you can hold for an hour' heuristic, because it's defined by the underlying steady-state physiology rather than a clock.

CP and FTP are close but not identical. Because CP is a steady-state asymptote, it typically corresponds to power a rider can hold for roughly 40–60 minutes, so it often lands a few percent above a 20-minute-test FTP estimate — how far depends on the test durations used and the size of the rider's W'. The classic estimation mistakes are fitting the model to efforts that are all too short (which inflates CP), pairing durations that are too close together, or not going genuinely maximal on each test effort.