Two main hypotheses compete to explain global cooling and the abrupt growth of the Antarctic ice sheet across the Eocene–Oligocene transition about 34 million years ago: thermal isolation of Antarctica due to southern ocean gateway opening1, 2, 3, 4, and declining atmospheric CO2 (refs 5,6). Increases in ocean thermal stratification and circulation in proxies across the Eocene–Oligocene transition have been interpreted as a unique signature of gateway opening2, 4, but at present both mechanisms remain possible. Here, using a coupled ocean–atmosphere model, we show that the rise of Antarctic glaciation, rather than altered palaeogeography, is best able to explain the observed oceanographic changes. We find that growth of the Antarctic ice sheet caused enhanced northward transport of Antarctic intermediate water and invigorated the formation of Antarctic bottom water, fundamentally reorganizing ocean circulation. Conversely, gateway openings had much less impact on ocean thermal stratification and circulation. Our results support available evidence that CO2 drawdown—not gateway opening—caused Antarctic ice sheet growth, and further show that these feedbacks in turn altered ocean circulation. The precise timing and rate of glaciation, and thus its impacts on ocean circulation, reflect the balance between potentially positive feedbacks (increases in sea ice extent and enhanced primary productivity) and negative feedbacks (stronger southward heat transport and localized high-latitude warming). The Antarctic ice sheet had a complex, dynamic role in ocean circulation and heat fluxes during its initiation, and these processes are likely to operate in the future.