El Niño is the largest year-to-year fluctuation of Earth’s climate, and its strongest events disrupt rainfall across much of the planet. Each event weakens the Pacific trade winds, which are maintained by the west–east sea-surface temperature (SST) contrast that El Niño itself erodes. Whether this relationship imposes an upper limit on event amplitude remains unknown. Across three SST reconstructions for 1950–2025, we find that the zonal SST contrast decreases by about 1°C per degree of Niño3.4 warming within each calendar month. Extrapolating this relationship to zero contrast gives a seasonal threshold ranging from +1.8°C in April to +4.3°C in November; no month since 1950 has exceeded its corresponding threshold. Relative to the tropical mean SST, the threshold has changed little since 1870, as strengthening of the climatological gradient and tropical warming nearly offset each other. In historical near-zero-gradient months, the ocean advective terms associated with the Bjerknes feedback reversed, although these cases occurred only during the decay season. Across 57 coupled climate models, a near-zero gradient during September–December is associated with a lower probability of further Niño3.4 growth: 29%, compared with 62% when 1–2°C of gradient remains. In 2026, the gradient reached a record September low while El Niño was still developing. We define prospective observational tests to assess whether the event continues to grow as the gradient approaches zero, providing a direct test of the proposed constraint.