The Stack and the Sovereignty of Computation
Benjamin Bratton's The Stack is a difficult but unusually useful book for the AI era because it treats computation as a planetary political architecture: mineral extraction, cloud platforms, cities, addresses, interfaces, and users arranged into a governing machine no single institution fully controls.
The Book
The Stack: On Software and Sovereignty was published by MIT Press in 2016 in its Software Studies series. MIT Press describes it as a political and design theory of planetary-scale computation, built around the claim that smart grids, cloud platforms, mobile apps, smart cities, Internet of Things systems, and automation should be read together as one accidental megastructure.
The book's value is that it refuses the small frame. It does not treat a platform as merely a company, an interface as merely a screen, or a user as merely a person clicking through software. It asks what kind of political geography appears when computation becomes infrastructure, jurisdiction, address system, labor system, archive, and perception layer at once.
That question is now more urgent than it was in 2016. Generative AI did not replace Bratton's model; it made the model easier to see. The AI boom ties together energy, chips, data centers, model platforms, city planning, identity systems, workplace software, companion interfaces, and new classes of synthetic users.
The Six Layers
Bratton organizes the Stack through six layers: Earth, Cloud, City, Address, Interface, and User. The ordering matters. Computation begins in matter: minerals, energy, water, land, cables, logistics, and climate modeling. It then condenses into cloud platforms, reorganizes cities, assigns addresses, mediates perception through interfaces, and produces users that may be human, institutional, machinic, or hybrid.
This is a useful corrective to the usual story of digital life. The internet is not weightless. AI is not a mind floating above the world. A model answer depends on mines, fabs, grids, fiber, data centers, platform contracts, software standards, policy decisions, interface defaults, and users trained to ask questions in forms the machine can accept.
The layer model is not a perfect map. It is better understood as a diagnostic scaffold. It lets a reader ask where power is being routed. Is the fight at the Earth layer, over electricity and materials? At the Cloud layer, over model access and compute? At the Address layer, over identity and permission? At the Interface layer, over what a person can perceive or contest?
Software Sovereignty
The book's core political move is to shift sovereignty away from the state alone. Bratton is not saying states disappear. He is saying that states, platforms, cities, protocols, databases, and users now overlap in ways that make older political maps insufficient.
A cloud provider may host the tools through which a public agency operates. A platform may set speech rules across borders. A phone may become an identity document, wallet, workplace, confession booth, map, classroom, and sensor package. A model provider may define what forms of knowledge can be searched, summarized, generated, or refused.
In that world, governance is not only law after the fact. It is also architecture before the fact. Defaults, APIs, app-store rules, procurement contracts, data schemas, model policies, moderation queues, and hardware bottlenecks become constitutional material. They shape what action is possible before anyone names the arrangement as politics.
The AI Stack
Bratton's 2026 MIT Press Reader essay, adapted from the tenth-anniversary edition, argues that AI is changing every Stack layer. That is the right update. AI turns the Stack from a networked computation theory into a theory of planetary cognition.
At the Earth layer, AI intensifies the politics of chips, rare minerals, energy, cooling, and land. At the Cloud layer, foundation-model companies and hyperscalers become gatekeepers for capabilities that many institutions cannot reproduce. At the City layer, data centers, robotics, logistics, policing, transit, and service systems alter local political economy.
At the Address layer, identity and authorization become harder because agents, bots, synthetic media, and human-machine teams all need to be named, permitted, logged, and bounded. At the Interface layer, chat, voice, augmented reality, generated video, and automated copilots change what feels real and actionable. At the User layer, the user is no longer just a citizen-consumer. The user may be a corporation, a model, a robot, a household, a swarm of agents, or a person partially acting through automation.
Where the Book Is Hard to Use
The Stack is not an easy book. Ian Bogost's review in Critical Inquiry credits its ambition while also noting the burden of its abstraction and density. That criticism is fair. The book often asks the reader to hold architecture theory, political theology, software studies, platform economics, science fiction, and infrastructure analysis in the same sentence.
The cost is practical. A book that wants to serve as a design brief sometimes withholds the plain operational checklist a policymaker, city planner, technologist, or organizer might need. Readers should not expect a finished regulatory program. They should expect a powerful frame that still needs translation into institutional practice.
That limitation does not ruin the book. It clarifies how to use it. Read it as a pattern map for infrastructure power, not as a policy manual. Its best pages make visible the vertical dependencies that smoother technology writing tends to hide.
The Site Reading
For this site, The Stack is a book about reality becoming administrable through computation. A system does not need to be conscious to govern. It only needs to organize perception, identity, access, memory, incentives, and dependency at scale.
That makes the interface morally serious. The surface where a person meets a system is not just usability. It is where the larger architecture becomes intimate: the prompt box, the feed, the dashboard, the chatbot, the ranking, the map pin, the trust score, the account status, the generated summary. A planetary machine becomes local by asking one person to click, accept, appeal, confess, verify, or comply.
The practical lesson is to audit by layer. Any serious AI institution should be able to say where its materials come from, who controls its cloud dependency, what city and labor effects it creates, how it addresses people and agents, what its interfaces make visible or invisible, and what kinds of users it produces. Without that layered account, AI governance becomes a conversation about model behavior detached from the world that makes the model powerful.
Sources
- MIT Press, The Stack: On Software and Sovereignty.
- Oxford Academic / MIT Press Scholarship Online, The Stack: On Software and Sovereignty.
- Benjamin H. Bratton, MIT Press Reader, "Benjamin Bratton on Planetary Computation's Next Phase", adapted from the tenth-anniversary edition of The Stack.
- Ian Bogost, Critical Inquiry, review of The Stack, February 16, 2017.
- Jakub Wolak, Praktyka Teoretyczna, "Uziemienie. Filozoficzne marginalia do The Stack i O antropolizie Benjamina Brattona", 2021.
Book links are paid affiliate links. As an Amazon Associate I earn from qualifying purchases.
- Amazon, The Stack by Benjamin Bratton.