Techniques, systems, and articles of manufacture for reducing latencies through packet aggregation. A method includes computing an end-to-end latency for each of multiple data packets to reach a pre-determined destination, determining an aggregation scheme with a corresponding aggregation time for each of the multiple data packets, wherein the aggregation time encompasses the time needed to transmit each of the multiple data packets to an aggregation component and the time spent waiting to output the aggregation component to an output port, and wherein the aggregation time is less than the computed end-to-end latency for each of the multiple data packets, aggregating each of the multiple data packets in accordance with the aggregation scheme, and programming the aggregation component with the aggregation scheme.