PhoenixAI raises $80M to drive the development of agentic AI-ready database technology
by Kyt Dotson · SiliconANGLEPhoenixAI Inc., formerly known as CelerData, today announced it has raised $80 million in new funding to fuel the development of the company’s artificial intelligence-native database and expand governance for regulated industries.
Sky9 Capital led the Series B round, with participation from Atypical Ventures and Olive Technology Ventures, as well as previous investors.
As agentic AI has shifted from planning and prototyping to production, it has brought AI agents from experiment to large-scale patterns. Instead of development boxes and virtual machines running small workloads, swarms of AI agents now spray databases with thousands or millions of questions a second that modern data stacks strain under.
PhoenixAI built what it calls an agentic AI-ready analytical database prepared to serve this new era of agents and the scaling enterprise demand.
Transactional databases process individual operations, such as inserting rows, updating account balances and recording orders. They act to record one thing at a time, reliably, atomically and with durability and highly normalized data. Agents struggle here because these data formats are optimized for narrow operations and agents work with unstructured, conversational use cases that humans think in: “What were our top 10 customers by revenue growth over the last 90 days, broken out by product line?” That’s an analytical question spanning millions of rows across multiple tables.
Analytical databases allow the storage of and handling of complex questions across massive datasets. They scan billions of rows, aggregate, join and summarize, but they sacrifice write speed for read speed. Examples include Snowflake Inc., ClickHouse Inc., Apache Druid and Google LLC BigQuery – and PhoenixAI.
An analogy might be a research analyst pulling every transaction from the last quarter, grouping by region, and calculating trends. You don’t need instant writes; you need fast, complex reads. These sit behind dashboards, drive business intelligence tools and now AI agents when they need to reason across massive enterprise datasets.
Analytical databases do not replace transactional databases; they sit alongside them. The transactional database acts as a system of record, acting as the source of truth, while the analytical database provides the system of insight for the agentic world. PhoenixAI isn’t here to replace a company’s Oracle or SAP enterprise resource planning system; it’s a layer that sits on top of it and makes it smarter so that AI agents can act and think faster.
PhoenixAI said it has rebuilt its analytical database to handle not just the agentic AI era but thousands of agent swarms looking for information.
“Most of today’s analytical databases were architected for a world that no longer exists, where humans ran dashboards on flat tables and complexity was someone else’s problem,” said President Rick Underwood. “When thousands of agents need to query, reason and act on petabytes of live data simultaneously — any question, simple or complex — the database is either the bottleneck or the breakthrough.”
PhoenixAI claims subsecond latency and high concurrency on live data, enabling multiple agents to query it simultaneously. This allows agents to ingest data rapidly while it’s being updated across massive spans – no more waiting, blockers or bottlenecks. The company calls this shape “no pipeline,” where fresh data is constantly rolling in from Kafka, an open-source event streaming platform that decouples data pipelines, to bring information up to date in seconds rather than minutes or hours.
The other major players in the analytical database market are not standing still either. Snowflake just launched its own agentic features. Databricks Inc. is pushing real-time with Delta Live Tables. ClickHouse Cloud has improved concurrency significantly.
The race is on to solidify the market category of “agentic database” and build out the infrastructure data layers that will feed this new hungry AI future. One where queries depend on more than just which database row to check and look at how analysis interacts with information.