In general, the use of adaptive services and adaptation frameworks for services is justified by the need of providing a flexible environment for service execution in changing environments, due to changes of context or changes in requirements. Such flexibility allows providing services in ways that are most suited to the current situation of invocation of the service. A parameter that is commonly used to assess the fitness of services in changing situations is quality of service, that commonly includes parameters such as, for instance, response time, availability, trust. Adaptivity can also be the basis for building dynamic service compositions driven by other types of goals, and in this talk the focus is on building adaptive services with the goal of improving energy efficiency. Energy efficiency is defined as the ability of a system to make an efficient use of the available resources. In variable and changing contexts the use of resources might be overprovisioned, in order to be able to cope with situations of system overload, maintaining the quality of service guarantees associated with a given service. As a result, in general we see a tradeoff between requirements imposed by quality of service and energy efficiency requirements. Adaptivity can help smoothing this tradeoff, since the services can be configured dynamically to exploit the available resources in a better way. We analyze energy efficiency in service compositions from two different perspectives. The first case is the execution of services in large service centers, in which services are executed dynamically sharing computing resources and storage systems. Such a case is studied in the GAMES (Green Active Management of IT Services) European project, in which IT resources are managed dynamically according to the context of execution and the characteristics of the services, which are driving adaptation policies. A second perspective is the use of dynamic services as enablers of energy efficiency strategies in given application domains, such as services in smart environments, e.g. in homes or buildings. In this case adaptive services can help reducing CO2 emissions since the energy consuming resources can be controlled by adaptive services, based on the context which is providing information about the environment and behavior of inhabitants. Research directions in both perspectives require the ability to manage monitoring information dynamically, to ensure an adequate level of granularity of the information, and to model and control the components of the environment in order to provide adaptivity to support energy efficiency on one hand, and on the other hand to guarantee the quality of service required by the applications.