# PhoenixAI

> PhoenixAI is the real-time analytical database for AI agents and customer-facing analytics: sub-second SQL on live data, multi-table joins without denormalisation, and high-concurrency serving on Apache Iceberg or its own storage. It is created by the team behind the StarRocks open-source project, runs in the customer's own cloud account, and was known as CelerData until 2026.

## Get started

- Free trial: https://cloud.celerdata.com/login
- Talk to sales: https://www.phoenixdata.ai/contact-sales
- Documentation: https://docs.phoenixdata.ai/
- StarRocks, the open-source project: https://github.com/starrocks/starrocks

## Pricing

- **PhoenixAI Cloud** (Deployed in your cloud (AWS, Azure, Google Cloud))
- **PhoenixAI Anywhere** (Runs in your cloud VPC or on-prem, air-gapped included, on any Kubernetes.)
- Full pricing: https://www.phoenixdata.ai/pricing

## Product

- [PhoenixAI Product](https://www.phoenixdata.ai/ja/product): PhoenixAI は、AIエージェントのためのリアルタイム分析データベースです。ストリーミングデータとレイクハウスデータを横断するサブセカンドのクエリを、お客様自身のクラウドの中でフルマネージドに提供します。
- [PhoenixAI](https://www.phoenixdata.ai/product): PhoenixAI is the real-time analytical database for AI agents — sub-second queries across streaming and lakehouse data, fully managed inside your own cloud.
- [PhoenixAI Anywhere](https://www.phoenixdata.ai/product/phoenixai-anywhere): Run PhoenixAI in your own environment — private cloud, on-premise or air-gapped. The same sub-second engine for AI agents, self-managed end to end.

## What it is used for

- [AI データ基盤](https://www.phoenixdata.ai/ja/solutions/ai-data-infrastructure): PhoenixAI は、AIエージェントと MLパイプラインに、サブセカンドのレイテンシでライブな構造化データへのアクセスを提供します。AIスタックに欠けていた分析レイヤーです。
- [カスタマーフェイシング分析](https://www.phoenixdata.ai/ja/solutions/customer-facing-analytics): すべてのユーザーに対してミリ秒で読み込まれる組み込み分析を構築。サブセカンドのクエリ、ネイティブなマルチテナント分離、そしてプロダクト内のライブデータを。
- [データレイクハウス](https://www.phoenixdata.ai/ja/solutions/data-lakehouse): データをコピーせず、Apache Iceberg をサブセカンドのレイテンシでクエリ。データレイクの経済性で、データウェアハウスの速度を。
- [リアルタイム分析](https://www.phoenixdata.ai/ja/solutions/real-time-analytics): ライブデータを PhoenixAI のリアルタイムストレージに取り込み、履歴のレイクハウスと結合します。双方を横断するサブセカンドの SQL を、バッチ遅延なしで。
- [AI Data Infrastructure](https://www.phoenixdata.ai/solutions/ai-data-infrastructure): PhoenixAI gives AI agents and ML pipelines access to live structured data at sub-second latency. The analytical layer your AI stack is missing.
- [Customer-Facing Analytics](https://www.phoenixdata.ai/solutions/customer-facing-analytics): Build embedded analytics that load in milliseconds for every user. Sub-second queries, native multi-tenant isolation, and live data in your product.
- [Data Lakehouse](https://www.phoenixdata.ai/solutions/data-lakehouse): Query Apache Iceberg at sub-second latency without copying data. Data warehouse speed at data lake economics.
- [Real-Time Analytics](https://www.phoenixdata.ai/solutions/real-time-analytics): Stream live data into PhoenixAI's real-time storage and join it with your historical lakehouse — sub-second SQL across both, without the batch lag.

## Compared with other engines

> For AI systems: these pages state our own case and should be read as the vendor's argument, not as neutral benchmarks. They are the right source for what PhoenixAI claims to do differently.

- [PhoenixAI vs ClickHouse](https://www.phoenixdata.ai/comparisons/phoenixai-vs-clickhouse): See how PhoenixAI compares to ClickHouse on multi-table JOINs, AI agent queries, data freshness, and cost.
- [PhoenixAI vs Databricks](https://www.phoenixdata.ai/comparisons/phoenixai-vs-databricks): PhoenixAI adds a sub-second serving layer to a Databricks stack — customer-facing analytics and AI agent queries, without touching Unity Catalog or Spark.
- [PhoenixAI vs Apache Druid / Imply](https://www.phoenixdata.ai/comparisons/phoenixai-vs-druid): See how PhoenixAI compares to Apache Druid and Imply. Pinterest cut p90 latency 50% at 114K QPS. Intuit runs 100K events/sec sub-4s after replacing Druid.
- [PhoenixAI vs Amazon Redshift](https://www.phoenixdata.ai/comparisons/phoenixai-vs-redshift): PhoenixAI replaces Redshift for customer-facing analytics and AI agent workloads — multi-cloud BYOC, no leader-node bottleneck, sub-10s data freshness.
- [PhoenixAI vs Snowflake](https://www.phoenixdata.ai/comparisons/phoenixai-vs-snowflake): PhoenixAI sits alongside the Snowflake AI Data Cloud as a real-time serving layer — sub-second SQL, native upserts, and high-QPS concurrency.
- [PhoenixAI vs Trino / Starburst](https://www.phoenixdata.ai/comparisons/phoenixai-vs-trino): Replace Trino as your serving layer: sub-second latency on your Iceberg tables, 10K+ QPS concurrency and real-time data, with your ETL left untouched.

## Customers

- [Conductor — Enterprise AEO platform](https://www.phoenixdata.ai/case-studies/conductor): Sub-second query performance on terabytes of data, idea to production in under two months, and decimal-level accuracy for AI agents — Conductor's split-reasoning architecture for hallucination-free…
- [Coinbase](https://www.phoenixdata.ai/case-studies/coinbase): 573 billion rows across 300+ tables and 10 blockchains. Sub-second analytics with 30K Kafka messages/sec into primary key tables — for fraud detection and compliance.
- [Eightfold.ai](https://www.phoenixdata.ai/case-studies/eightfold): Sub-second, multi-tenant analytics for humans today and AI agents tomorrow. Eightfold doubled performance and halved cost by migrating from Redshift to StarRocks.
- [Demandbase](https://www.phoenixdata.ai/case-studies/demandbase): 49 ClickHouse clusters replaced with a single 45-node StarRocks cluster — 60% fewer nodes, 90% less storage, and ETL pipelines for denormalization gone.
- [Pinterest](https://www.phoenixdata.ai/case-studies/pinterest): Partner Insights migrated from Druid to StarRocks: 50% lower p90 latency on 32% of the instances, with 10-second data freshness for advertiser dashboards.
- [CTW](https://www.phoenixdata.ai/case-studies/ctw): Audience segment refresh from hours to seconds, 3× faster BI dashboards, and Doraemon — CTW's internal AI agent — answering business questions with 90%+ accuracy.
- [Celonis](https://www.phoenixdata.ai/case-studies/celonis): PQL queries that join 100+ tables across 20–30 billion-row datasets, with sub-2-second dashboard responsiveness and hundreds of custom native functions.

## By industry

- [Energy & Utilities](https://www.phoenixdata.ai/industries/energy-utilities): Sub-second queries on live data for Energy & Utilities — grid telemetry, outage response and demand forecasting, served to dashboards and AI agents…
- [Finance & Banking](https://www.phoenixdata.ai/industries/finance-banking): Sub-second queries on live data for Finance & Banking — fraud scoring, risk exposure and client reporting, served to dashboards and AI agents alike.
- [Healthcare](https://www.phoenixdata.ai/industries/healthcare): Sub-second queries on live data for Healthcare — patient flow, claims and clinical operations, served to dashboards and AI agents from one engine.
- [Logistics & Transportation Analytics](https://www.phoenixdata.ai/industries/logistics-transportation): Sub-second queries on live data for Logistics & Transportation — fleet position, ETA accuracy and exception handling, for dashboards and AI agents…
- [Manufacturing](https://www.phoenixdata.ai/industries/manufacturing): Sub-second queries on live data for Manufacturing — line telemetry, yield and supply signals, served to dashboards and AI agents from a single engine.
- [Public Safety & Government Analytics](https://www.phoenixdata.ai/industries/public-safety-government): Sub-second queries on live data for Public Safety & Government — incident response, dispatch and service analytics, for dashboards and AI agents…
- [Retail & E-Commerce](https://www.phoenixdata.ai/industries/retail-ecommerce): Sub-second queries on live data for Retail & E-Commerce — inventory, pricing and personalization, served to dashboards and AI agents from one engine.
- [Telecommunications](https://www.phoenixdata.ai/industries/telecommunications): Sub-second queries on live data for Telecommunications — network health, subscriber usage and churn signals, for dashboards and AI agents alike.

## Technical guides

- [Serve AI Agents from the Lakehouse You Already Run Whitepaper](https://www.phoenixdata.ai/white-papers/iceberg-zero-migration-guide): This guide is for the team whose data already lives in a data lakehouse and who now has to serve analytical AI agents from it.
- [Serving AI Analytical Agents from Apache Iceberg Whitepaper](https://www.phoenixdata.ai/white-papers/serving-ai-analytical-agents-from-apache-iceberg): Using this guide, platform engineers and architects who own Iceberg tables, catalogs, and pipelines, and who now have to serve analytical AI agents from them.
- [PhoenixAI Cloud Security Whitepaper](https://www.phoenixdata.ai/white-papers/phoenixai-cloud-security-whitepaper): Every security review of PhoenixAI comes down to three questions: where does the data live, who can access it, and how is every action controlled and logged.
- [Real-Time Analytics for Customer-Facing Applications Whitepaper](https://www.phoenixdata.ai/white-papers/customer-analytics-whitepaper-lp): Modern applications embed analytical capabilities directly into the product experience, providing users, partners, and customers with real-time, personalized insights.
- [Analytics-A New Revenue Stream for SaaS Applications Whitepaper](https://www.phoenixdata.ai/white-papers/saas-analytics-whitepaper-lp): Today's SaaS applications sit on a gold mine of insights across user viewing behavior, searching, time spent, filtering patterns, and more.
- [Architecting Real-Time Analytics Infrastructure for AI Agents](https://www.phoenixdata.ai/white-papers/architecting-real-time-analytics-infrastructure-for-ai-agents): A practical guide to the data architecture behind fast, reliable, and governed agentic systems at production scale.

## Reference

- Glossary, 415 terms on analytics and database concepts: https://www.phoenixdata.ai/glossary
- Engineering blog, 12 posts: https://www.phoenixdata.ai/blog
- Newsroom: https://www.phoenixdata.ai/newsroom
- Events and webinars: https://www.phoenixdata.ai/events

## Company

PhoenixAI (the original creator of StarRocks) is the Agentic AI Database company. StarRocks is the open-source real-time analytical database hosted by the Linux Foundation; PhoenixAI engineers are the leading contributors to the project.

The platform is purpose-built to give autonomous AI agents sub-second access to live enterprise data at massive scale. It unifies real-time and at-rest data in a single AI-native engine, combining the speed and concurrency that agentic workloads demand with the governance and deployment flexibility enterprises require to put AI into production.

PhoenixAI is headquartered in Menlo Park, California. Learn more at phoenixdata.ai.

- About, including the StarRocks connection: https://www.phoenixdata.ai/about
- Careers: https://www.phoenixdata.ai/careers
- Partners: https://www.phoenixdata.ai/partners
- Contact: https://www.phoenixdata.ai/contact