A decentralized data architecture where domain teams own and serve their data as products.
Data Mesh is a sociotechnical approach to data architecture that decentralizes data ownership and treats data as a product. Rather than centralizing all data in a single data warehouse or lake managed by a central data team, Data Mesh distributes data ownership to domain teams — the people closest to the data — who are responsible for serving their data as high-quality, discoverable products to the rest of the organization.
Data Mesh has become the dominant architectural philosophy for large-scale enterprise data in 2026, but its success depends critically on semantic interoperability. When dozens of domain teams each serve their own data products, the risk of semantic fragmentation is high — each team may define the same concepts differently. Semantic layers and shared ontologies are the governance mechanisms that make Data Mesh work at scale without creating a semantic Tower of Babel.
Data Mesh is built on four principles: domain-oriented decentralized data ownership, data as a product (discoverable, addressable, trustworthy, self-describing), self-serve data infrastructure (platforms that enable domain teams to build data products without central bottlenecks), and federated computational governance (global standards enforced locally). The semantic layer provides the federated governance layer that ensures semantic consistency across domain data products.
A large e-commerce company implements Data Mesh. The Orders domain team owns and serves order data as a product; the Customers team owns customer data; the Inventory team owns inventory data. Each team publishes their data product with a semantic contract — a formal definition of all metrics and entities using the company's shared ontology. AI agents can query across domain data products knowing that 'customer' means the same thing in every domain.