The standard picture of power that governs most political thinking goes roughly like this: power is held by some people and not others. Those who hold it use it to constrain, coerce, or exploit those who do not. The state has power; citizens are subject to it. The king has power; peasants are subject to it. The corporation has power; workers are subject to it. Power flows downward from a center. The proper political response is to redistribute it more fairly, to constrain it through rights and institutions, or in the more radical version, to seize it from those who wrongly hold it.
Michel Foucault (1926–1984) spent his career dismantling this picture, not because he thought power was not a problem but because he thought the standard picture was so inadequate to how power actually works in modern societies that political thinking organized around it would consistently misidentify what needed to be contested and where.
His alternative is not a theory in the systematic philosophical sense. It is a series of genealogical investigations, detailed historical studies of how specific practices, institutions, and discourses emerged, changed, and produced specific kinds of subjects and specific forms of knowledge. He studied the history of madness, the history of clinical medicine, the history of the prison, the history of sexuality, the history of the human sciences. In each case, his method was not to explain why a practice emerged (what interests it served, whose rationality drove it) but to trace its contingent historical emergence: how things could have been otherwise, how what appears natural and inevitable is actually the product of specific historical conjunctures.
As the SEP notes, Foucault began talking about power as soon as he began doing genealogy, because what the genealogical method revealed was that knowledge and power are not separate. The development of clinical medicine did not just produce knowledge about disease and health; it produced a new form of surveillance of the body, a new ability for institutions to monitor, classify, and normalize individuals. The development of psychiatry did not just produce knowledge about mental illness; it produced the distinction between the sane and the insane as a social technology for managing populations. The development of criminology and the prison system did not just produce knowledge about crime; it produced the delinquent as a social type, whose identification and management extended disciplinary power throughout society.
The concept that captures this is power/knowledge (pouvoir/savoir). Power and knowledge are not two separate things that happen to be related. They are mutually constitutive: power requires and produces knowledge (surveillance produces data; data enables more precise control), and knowledge is always also a form of power (naming, classifying, and diagnosing is itself an act of control). There is no innocent knowledge, no view from nowhere, no scientific expertise that stands outside the power relations in which it operates. This does not mean that all knowledge claims are equally valid or that science is just politics. It means that the conditions under which knowledge is produced, the social functions it serves, and the subjects it constitutes are always implicated in power relations, whether the knowledge-producers acknowledge this or not.
Quick reflection
Think about an institution you have been part of — a school, a hospital, a workplace, a prison if you have encountered one. What kind of knowledge about you did that institution produce? Who had access to it? How was it used to classify, manage, or normalize your behavior? Now ask: was this knowledge about you primarily accurate (did the institution know something real about you) or was it a construction (did the institution's categories shape what you became in a way that would not have happened without them)? What is the difference, and does Foucault help you think about it?