In this note, we study large deviations of the number N of intercalates (2 × 2 combinatorial subsquares which are themselves Latin squares) in a random n × n Latin square. In particular, for constant δ > 0 we prove that Pr(N ≤ (1 − δ)n 2 /4) ≤ exp(−Ω(n 2 )) and Pr(N ≥ (1 + δ)n 2 /4) ≤ exp(−Ω(n 4/3 (log n) 2/3 )), both of which are sharp up to logarithmic factors in their exponents. As a consequence, we deduce that a typical order-n Latin square has (1 + o(1))n 2 /4 intercalates, matching a lower bound due to Kwan and Sudakov and resolving an old conjecture of McKay and Wanless. Kwan was supported by NSF grant DMS-1953990. Sah and Sawhney were supported by NSF Graduate Research Fellowship Program DGE-1745302.1 To see the analogy to permutation matrices, note that a Latin square can equivalently, and more symmetrically, be viewed as an n × n × n zero-one array such that every axis-aligned line sums to exactly 1.2 Jacobson and Matthews [30] and Pittenger [48] designed Markov chains that converge to the uniform distribution, but it is not known whether these Markov chains mix rapidly.