The insertion of an olefin into a preformed metal–carbon bond is a common mechanism for transition‐metal‐catalyzed olefin polymerization. However, in one important industrial catalyst, the Phillips catalyst, a metal–carbon bond is not present in the precatalyst. The Phillips catalyst, CrO3 dispersed on silica, polymerizes ethylene without an activator. Despite 60 years of intensive research, the active sites and the way the first CrC bond is formed remain unknown. We synthesized well‐defined dinuclear CrII and CrIII sites on silica. Whereas the CrII material was a poor polymerization catalyst, the CrIII material was active. Poisoning studies showed that about 65 % of the CrIII sites were active, a far higher proportion than typically observed for the Phillips catalyst. Examination of the spent catalyst and isotope labeling experiments showed the formation of a Si–(μ‐OH)–CrIII species, consistent with an initiation mechanism involving the heterolytic activation of ethylene at CrIIIO bonds.