A robot does not spend its energy only on computing. Total task energy is compute plus actuation, and both terms are measured. On a reaching task the compute share runs from five percent at one watt of compute to fifty-nine percent at thirty watts, and on a mobile robot under autonomous navigation the graphics processor alone draws thirty-seven percent while the motors draw sixteen. Improve compute by any factor and the actuation term does not move, so the total falls to the non-compute share plus the compute share divided by that factor, and the limit is the reciprocal of the non-compute share. It sits between about one and a half and two and a half times. A ten-thousandfold compute improvement is worth two and a half times in this denominator, and a thirty-fold improvement already delivers ninety-five percent of that. This is a property of the denominator rather than of any substrate. It tells you two useful things: where further compute work stops paying in this denominator, so effort moves to the actuation term, and which quantities to develop the embodied case in. Latency against a control period, duty cycle between tasks, and thermal envelope are set by different physics and carry no limit of this shape.