The energy‐saving potential of the internal thermally coupled air separation column (ITCASC) is well‐established, but distinct dynamic characteristics and control loop interactions make it inflexible to control. To take care of high‐purity ITCASC control complications, a state‐space model predictive control (MPC) was formulated. A direct finite‐horizon control approach was exploited to align the dynamic states with the model predictions. MPC‐I and MPC‐II were developed, and further compared to a previous adaptive multivariable generalized prediction control (AM‐GPC). The results obtained show that the control performance of the proposed MPC‐II is superior to that of MPC‐I and AM‐GPC.