This paper investigates short-term torque capability maximization in electric machines based on thermal time ratings. To understand time ratings, machine transient thermal response is divided into four regimes: subtransient, transition, transient, and temperature creep. Each regime is discussed separately. For modest overload conditions, an exponential characteristic based on limited geometric data is formulated for the time ratings, while for severe overload cases, linear temperature rise linked to known 'i-squared-t' (i 2t) conductor rating approaches suffice to characterize the temperature rise. It is shown that both exponential and i2t methods fail to provide accurate estimates for moderate overload operation. In such cases heat transfer inside the machine is solved analytically to obtain time ratings. Experiments with a 5 hp induction machine were conducted to evaluate the thermal characteristics and insulation material properties. This machine is used to validate the theoretical foundation presented for time ratings. The results show that time rating estimation based on limited geometric data is accurate within an error range of about 7%. Implementation of these time ratings for real-time torque maximization is discussed.