PhoenixAI leaves legacy analytics ashes behind to build AI agent database. Published mon 15 Jun 2026 // 14:04 UTC Agentic AI database startup PhoenixAI has raised $80 million in a B-round to develop its SW and expand its go-to-market initiatives. PhoenixAI rebranded from CelerData last month. That company, founded in 2022, developed a real-time and unified analytics engine using open-source, massively parallel StarRocks code. The three CelerData co-founders; CEO James Li, COO Andy Ye, and CTO Alvin Zhao co-founded the StarRocks venture in 2020. StarRocks is focussed on high-performance SQL queries, streaming ingestion and sub-second analytics at massive scale. Zhao says "over 300 leading companies worldwide have used StarRocks to build next-gen data analytics capabilities, including Airbnb, Celonis, Fanatics, Tencent, Trip.com, SF Technology, Didi, JD.com, and ZhongAn Insurance. Thousands of StarRocks nodes are running stably in customers' production environments." Rick Underwood, the ex-CEO of Commvault-acquired Clumio, joined CelerData as its President in January, leading go-to-market and operations. He said: "Today's agentic landscape has moved quickly from planning and prototyping to full-on production for mission-critical work - serving customers, managing supply chains, and driving internal workflows. "Agents now fire off thousands of unplanned, real-time queries, often swarming systems with questions that weren't anticipated when the data stack was designed, demanding analysis across live data, historical records, and multiple systems at once, which strains existing data stacks. PhoenixAI closes that gap by giving enterprises a faster, more efficient way to serve those workloads at scale with the level of governance the C-suite expects." As of 2024, CelerData provided high-performance data lakehouse analytics SW, using the StarRocks SQL query engine, with on-premises CelerData Enterprise software and the CelerData Cloud managed service. It carried on developing this, but then saw that AI agents were coming. These agents would substantially change the data warehouse/lakehouse query/response environment Phoenix says AI "agents query differently than people: they operate in real time, issuing unpredictable questions no historical analytics stack was built to handle. As a result, answering agent questions requires joining live and historical data on the fly in ways that traditional databases were never designed to support. Traditional databases avoid that work by requiring teams to pre-model and reshape data for each "expected" question that human users may ask, but that approach breaks down the moment agents start asking new questions at scale." Its SW PhoenixAI is built around three themes: speed, simplicity, and governance. Speed means sub-second latency and high concurrency on live data. Simplicity means a unified engine for real-time and historical data, eliminating separate pipelines and reducing total cost of ownership. Governance means built-in data sovereignty, role-based access control, SOC 2 standards, and encryption, deployed in the customer's own cloud (BYOC) or self-hosted with PhoenixAI Anywhere. PhoenixAI has a PhoenixAI Custom Connector for Claude, available for PhoenixAI Cloud customers, which can connect Claude directly to PhoenixAI and start asking questions in plain English, with results streaming back into the chat. Menlo Park headquartered PhoenixAI already has customers using its SW in production, such as AppLovin, Coinbase, Conductor, and Demandbase. We understand they were CelerData StarRocks users before the PhoenixAI rebranding. Xinyu Liu, a Senior Staff Software Engineer at, Coinbase, provided a characteristic quote: "PhoenixAI changed the equation: streaming updates from Kafka become queryable within seconds, analysts get sub-second responses on live normalized data, and our AI agents operate on the same real-time dataset. This level of performance at scale fundamentally changes what data teams can do." Ryan Nowacoski, Senior Engineering Manager, Data Platform, at Demandbase provided another: "PhoenixAI gives us a fast, isolated warehouse for agent workloads directly on our Apache Iceberg tables, with the optimizer handling novel joins automatically." The B-round was led by Sky9 Capital, with participation from Atypical Ventures and Olive Technology Ventures, and previous investors. Sky9 Capital is a global venture firm investing in AI, fintech, and deep tech, with offices in San Francisco, Boston, Beijing, Shanghai, and Singapore. Although there is no publicly recorded funding for CelerData, the B-round term implies a previous A-round, as does the "previous investors" point. Underwood has said on LinkedIn: "PhoenixAI is the company behind Linux Foundation StarRocks and we're becoming the Agentic AI Database. Every enterprise will soon have thousands of AI agents querying analytical data in real time. The database underneath has to be sub-second, massively concurrent, and open. That's what we've built. "I came here because this felt like the beginning of something transformative the same energy I felt walking into Snowflake in 2017. As President, I lead the business: go-to-market, finance, HR and operations. My co-founders built the technology; I'm building the company around it." Bootnote. We're told that, earlier in Andy Ye's career, he spent several years at Umeng in senior engineering and leadership roles. Umeng was then acquired by Alibaba at the end of 2013. He then served as a Director at Alibaba Group, later became Chief Product Officer at Shanghai Xforceplus, and subsequently co-founded StarRocks in 2020 and PhoenixAI (formerly CelerData) in 2022, where he has served as Co-Founder & COO and now Co-Founder & Chief Product Officer. There were earlier outside investments across the seed and Series A stages from a mix of institutional investors, including Amino Capital and the StarRocks open-source entity, prior to the current round of funding.

PhoenixAI, a database company built for AI agents, has raised $80 million in a Series B round led by Sky9 Capital, with participation from Atypical Ventures, Olive Technology Ventures and existing investors. The company addresses challenges traditional databases face when serving AI agents, which issue unpredictable, real-time queries across live and historical data. PhoenixAI's AI-native database delivers sub-second responses to agent queries whilst maintaining enterprise governance standards. Companies including AppLovin, Coinbase, Conductor and Demandbase are already using PhoenixAI in production. The funding will accelerate product development, expand go-to-market initiatives and deepen governance capabilities for regulated industries. Rick Underwood, previously CEO of Clumio and a senior leader at Snowflake, serves as president. PhoenixAI, formerly CelerData, is headquartered in Menlo Park, California.

PhoenixAI, formerly CelerData, has relaunched as an "Agentic AI Database" designed to handle queries from autonomous AI agents at sub-second latency and massive scale. The company has appointed Rick Underwood as president to lead its growth phase. The San Francisco-based startup addresses the challenge of AI agents querying enterprise data differently than human analysts, requiring thousands of concurrent queries with sub-second latency on live data. Underwood, previously CEO of Clumio, held senior roles at Snowflake through its 2020 IPO and led sales at Wavefront and Semmle before their acquisitions. PhoenixAI's platform unifies real-time and historical data whilst providing governance features including data sovereignty, role-based access control and SOC 2 compliance. It's available through bring-your-own-cloud deployment or self-hosted options.