PhoenixAI Cloud (StarRocks)

PhoenixAI, formerly known as CelerData, is a company headquartered in Menlo Park, California that created StarRocks, the open-source real-time analytical database hosted by the Linux Foundation under the Apache 2.0 license, and still leads its development. It targets teams that need fast SQL analytics on fresh data for dashboards, customer-facing analytics and AI agents.

The platform runs arbitrary joins without precomputation, makes updates from Kafka and Flink CDC queryable within seconds and queries live tables together with Apache Iceberg history in one SQL layer. PhoenixAI Cloud is fully managed inside the customer's AWS, Azure or GCP account, while PhoenixAI Anywhere is self-managed on Kubernetes, in a VPC, a data center or an air-gapped site. Isolated warehouses separate agent, BI, ETL and ad-hoc workloads. Users include Coinbase, Pinterest, Expedia, Tencent and miHoYo.

• Real-time ingestion: Updates from Kafka and Flink CDC are queryable within seconds.
• Join without precomputation: Arbitrary multi-table joins run without denormalization.
• Lakehouse queries: Unified SQL over live tables and Iceberg, Delta and Hudi data on S3, GCS or ADLS.
• Isolated warehouses: Separate compute for agent, BI, ETL and ad-hoc workloads.
• Flexible deployment: Managed in the customer's cloud account or self-managed, including air-gapped sites.
• Observability: Query insights plus live ingest and latency monitoring.

A Vietnamese e-commerce or ride-hailing company can power real-time merchant dashboards directly from Kafka streams. A bank can query an Iceberg lakehouse for risk and marketing analytics without copying data into a separate warehouse. A game publisher can analyze player events in near real time for live operations.

Data from Kafka and Flink CDC becomes queryable in under five seconds, and the vendor cites sub-second p95 latency at more than 1,000 queries per second. Querying Iceberg, Delta and Hudi tables directly reduces copies of data. Published case studies report Demandbase cutting hardware by 60% and storage cost by 90% while consolidating 49 clusters into one, and Coinbase serving 573 billion rows with sub-second latency.

Companies are moving from closed data warehouses to open lakehouse formats such as Apache Iceberg, and need a fast query layer on top. Customer-facing analytics and AI agents also demand sub-second answers on fresh data. Gaming, fintech and e-commerce firms across Asia already use StarRocks for these needs.

Formerly CelerData, headquartered in Menlo Park, California. Raised an $80 million Series B led by Sky9 Capital with Atypical Ventures and Olive Technology Ventures, announced June 11, 2026. Creator of StarRocks, hosted by the Linux Foundation. Partners include AWS, Confluent and Databricks.