Enhanced alpha pow­er com­pared with a base­line can reflect states of increased cog­ni­tive load, for exam­ple, when lis­ten­ing to speech in noise. Can knowl­edge about “when” to lis­ten (tem­po­ral expec­ta­tions) poten­tial­ly coun­ter­act cog­ni­tive load and con­comi­tant­ly reduce alpha? The cur­rent mag­ne­toen­cephalog­ra­phy (MEG) exper­i­ment induced cog­ni­tive load using an audi­to­ry delayed-match­ing-to-sam­ple task with 2 syl­la­bles S1 and S2 pre­sent­ed in speech-shaped noise. Tem­po­ral expec­ta­tion about the occur­rence of S1 was manip­u­lat­ed in 3 dif­fer­ent cue con­di­tions: “Neu­tral” (unin­for­ma­tive about forepe­ri­od), “ear­ly-cued” (short forepe­ri­od), and “late-cued” (long forepe­ri­od). Alpha pow­er through­out the tri­al was high­est when the cue was unin­for­ma­tive about the onset time of S1 (neu­tral) and low­est for the late-cued con­di­tion. This alpha-reduc­ing effect of late com­pared with neu­tral cues was most evi­dent dur­ing mem­o­ry reten­tion in noise and orig­i­nat­ed pri­mar­i­ly in the right insu­la. More­over, indi­vid­ual alpha effects dur­ing reten­tion account­ed best for observed indi­vid­ual per­for­mance dif­fer­ences between late-cued and neu­tral con­di­tions, indi­cat­ing a trade­off between allo­ca­tion of neur­al resources and the ben­e­fits drawn from tem­po­ral cues. Over­all, the results indi­cate that tem­po­ral expec­ta­tions can facil­i­tate the encod­ing of speech in noise, and con­comi­tant­ly reduce neur­al mark­ers of cog­ni­tive load.