# Google Distributed Cloud This is a hyperscaler's own platform, including its frontier models, packaged to run inside somebody else's building: a customer's data center, a government facility, a regulated institution, or a disconnected environment. It is one implementation of a pattern that every large cloud now offers, and the pattern matters more than any one product. > [!note] The one-sentence version > When the data cannot come to the cloud, the cloud is shipped to the data, with the frontier model inside it. ## Why the pattern exists A large set of institutions hold data they may not export: patient records, financial detail, classified or controlled information, material under residency law, and industrial process data of competitive value. Their alternative to forgoing the capability is to obtain it inside a boundary they control; see [[wiki/Private Model Hosting|Private Model Hosting]]. Meanwhile the providers have an interest in serving those institutions, which between them represent a very large share of consequential computing. The result is a category in which the provider's control plane, orchestration, security model, and increasingly its models are delivered as a self-contained system that runs on hardware in the customer's facility, managed either by the provider remotely or entirely by the customer in disconnected configurations. ## What it typically includes A hardware and software stack installed on premises, running the provider's container orchestration and managed services; see [[wiki/Kubernetes|Kubernetes]]. Accelerators for local inference and sometimes training. Model serving of the provider's frontier and open models locally, so prompts and outputs never leave the facility; see [[wiki/Inference Serving|Inference Serving]] and [[wiki/Inference Endpoint|Inference Endpoint]]. Retrieval grounded in the institution's own corpus; see [[wiki/Retrieval-Augmented Generation|Retrieval-Augmented Generation]]. Identity, policy, logging, and audit sufficient for the institution's compliance obligations. And, in the strictest configurations, fully disconnected operation updated by controlled media. ## The sovereignty dimension, stated carefully This category is routinely described as sovereign AI, and precision is worth insisting on. Running a provider's stack and models inside a national facility delivers real and valuable properties: data residency, physical control, audit, and continuity through a connectivity loss. It does not by itself deliver independence from the provider, because the models, the platform, and the supply chain remain the provider's. Both facts are true, and the distinction is exactly what [[wiki/Compute Sovereignty|Compute Sovereignty]] and [[wiki/Exit Cost|Exit Cost]] exist to measure: sovereignty is the practical ability to substitute, not the location of the rack. For the United States the pattern is straightforwardly advantageous. It lets American platforms and American models serve American hospitals, agencies, laboratories, and manufacturers under American law, inside their own walls, which is the mechanism by which frontier capability reaches institutions that could never adopt a public multi-tenant service; see [[wiki/Sovereign AI|Sovereign AI]] and [[wiki/National Compute Capacity|National Compute Capacity]]. Every large provider offers a version, which means the institutions in question have genuine choice among them. ## Relationships - **implements:** [[wiki/Sovereign AI|Sovereign AI]] · [[wiki/Private Model Hosting|Private Model Hosting]] - **delivered by:** [[wiki/Hyperscaler|Hyperscaler]] - **runs:** [[wiki/Kubernetes|Kubernetes]] · [[wiki/Inference Serving|Inference Serving]] · [[wiki/Retrieval-Augmented Generation|Retrieval-Augmented Generation]] - **measured against:** [[wiki/Compute Sovereignty|Compute Sovereignty]] · [[wiki/Exit Cost|Exit Cost]] - **secured by:** [[wiki/Confidential Computing|Confidential Computing]] · [[wiki/Data Center Security|Data Center Security]] - **part of:** [[wiki/Cloud Deployment|Cloud Deployment]] - **source articles:** [[articles/Vertically Integrating an AI Superpower from AI Factory to Citizen|Vertically Integrating an AI Superpower from AI Factory to Citizen]] · [[articles/The Gigawatt Is the New Measure of War|The Gigawatt Is the New Measure of War]] - **collection route:** [[collections/American Infrastructure|American Infrastructure]] · [[collections/American|American Collection]] ## Evidence posture - **Established:** the category of provider platforms operated inside customer facilities, including container orchestration, managed services, accelerators, local model serving, corpus-grounded retrieval, identity and audit, and disconnected configurations updated by controlled media; every large cloud provider offering a version. - **Established:** the institutional driver, namely data that may not be exported for regulatory, residency, classification, or competitive reasons. - **Strategic assessment:** the distinction between data residency with physical control and practical independence from a provider is an argued position, measured through substitution capability rather than facility location. ## Sources - Public distributed and on-premises cloud platform documentation, as of October 2026.