Can AI take care of patients? Is it possible to delegate care to machines? This contribution aims to answer these questions on a conceptual level, i.e. starting from the possible compatibility between the concept of “artificial intelligence” and that of “care”. Following the distinction between “curing” and “caring”, I will argue that, although AI is an effective support for curing (albeit with some ethical problems), it fails to achieve genuine caring. The paper will be divided into three parts. In the first, I will discuss the definition of AI as a split between intelligent thought and action (according to Luciano Floridi). I will then analyze the definition of care proposed by the ethics of care in particular by Nel Noddings as a relationship based on engrossment and motivational displacement, in which thought and action are fused. In part 3, I will consider my thesis, focusing mainly on the case of care robots, reconstructing the debate between scholars for and against the use of robots for care tasks. The main points on which the two sides confront each other are: (a) the heavy burden of care work, (b) the importance of the human touch, (c) the emotional investment in the machine, (d) the morality of deception in the care relationship. Although the debate is full of interesting arguments, there is a lack of conceptual analysis. I will argue that the two main traits of caring, i.e. being a practice and being a relationship, exclude the use of AI in principle. This does not mean that we cannot employ these machines in the health sector, even with some success, but that what the machine does cannot properly be called “caring” and that, therefore, we cannot completely delegate our work to machines except at the price of giving up this important aspect of our life together.