The performance of quantum heat engines is generally based on the analysis of a single cycle. We challenge this approach by showing that the total work performed by a quantum engine need not be proportional to the number of cycles. Furthermore, optimizing the engine over multiple cycles leads to the identification of scenarios with a quantum enhancement. We demonstrate our findings with a quantum Otto engine based on a two-level system as the working substance that supplies power to an external oscillator.PACS numbers: 03.65. Ta, 05.70.Ln Advances in technology have spurred the fabrication and study of thermal machines at the nanoscale, whose performance is governed by quantum fluctuations. Prominent examples include quantum heat engines (QHEs) and pumps [1][2][3][4]. Various prototypes have been realized in the laboratory by means of cold atoms and trapped ions as a working substance [5,6]. Theoretical studies of these machines are largely motivated by foundational questions that address the interplay between thermodynamics and statistical mechanics in the quantum world [7,8]. At the same time, exciting applications are in view. Processes varying from laser emission [1] to light harvesting in both artificial and natural systems [9-11] can be described in terms of QHEs.Nonetheless, the quest for quantum signatures of the performance of thermal devices remains challenging. It is understood that a universal behavior emerges in the limit of small action [12]. Identifying scenarios exhibiting quantum supremacy, with a performance surpassing that in classical thermodynamics, stands out as an open problem. To this end, the use of quantum coherence [13], nonequilibrium reservoirs [14,15], and many-particle effects [16,17] has been proposed.The performance of quantum thermal machines is usually assessed via the characterization of a single cycle, as in classical thermodynamics. This approach assumes that the average single-cycle efficiency and power carry over to an arbitrary number of cycles, i.e., work done through n cycles is expected to be equal to n times the work done per cycle. Yet, in quantum mechanics work is determined via projective energy measurements at the beginning and end of a prescribed protocol [18,19]. As a result, assessing the performance of a quantum thermal machine can severely alter its dynamics due to the quantum measurement backaction. We argue that the QHE performance can be best assessed by measurements on an external system on which work is done (see, e.g., [20] for a related discussion). By analyzing the dynamics over many cycles, we elucidate the role of the intercycle coherence and findSchematic quantum heat engine. The quantum engine E does work w on an external system S through the coupling H SE absorbing heat Q from the baths collectively represented by B, which consists of hot (B 1 ) and cold (B 2 ) baths.scenarios with quantum-enhanced performance. In particular, we demonstrate that the average amount of work through n cycles need not be proportional to n; rather, it may have an a...