Edelbyte exits stealth with $70 million for AI-agent data platform

Edelbyte's pitch rests on a shift in who asks the questions. Data platforms were built assuming humans, who can only type so many queries a day, so keeping things fast meant shortening histories and limiting queries to control cost. Sequoia argues AI agents investigating a problem may fire 50 queries where a person would run one, and that an agent's usefulness is decided by how much it can ask — which in turn depends on the data layer. The founders' background gives that argument a concrete reference point: Kusto was designed to return answers from high-volume log, telemetry and security event data in seconds rather than hours, and it became the foundation for Microsoft Sentinel, Microsoft Defender and Azure Monitor as well as the Azure Data Explorer product. Sequoia describes Edelbyte not as a rebuild of that engine but as a rethink of what a data platform should be for the AI era, bundling search, vector, graph and geospatial workloads into one system that answers in milliseconds at hundreds of petabytes, with a built-in knowledge base for agents. Whether that translates into durable demand is not settled by the announcement; the cited customer that ran 10 times more queries on the same budget is the kind of evidence the company will need more of.