[2/2] Databases in 2025: A Year in Review by Andy Pavlo (Category Databases)
Continuing my account of Andy Pavlo’s database market review, here are the remaining points.
💸 Money, mergers and acquisitions Pavlo describes a year full of large data-sector deals:
- The hunt for PostgreSQL companies continued: Neon went to Databricks and Crunchy Data to Snowflake, for $250 million.
- IBM acquired DataStax, known for Cassandra, for roughly $3 billion.
- Salesforce spent around $8 billion on the long-established ETL platform Informatica.
- MariaDB Corp. even managed to buy “itself” twice in one year, bringing the spun-off cloud subsidiary SkySQL back into the fold.
- A surprise was the Fivetran–dbt Labs merger: complementary businesses forming an ETL giant with an eye on an IPO.
📉 Investors cooled on new database systems Vector database hype faded, and venture investors put less money into young database startups in 2025, favouring AI instead. Established leaders still raised large rounds:
- Databricks raised $5 billion across two rounds during the year.
- ClickHouse raised $350 million.
- Supabase raised $300 million in total across Series D and E.
- Even young projects such as SpiralDB, the creator of Vortex, raised tens of millions.
💀 Not everyone survived Several companies did not make it through the year:
- FaunaDB closed. It was an ambitious distributed database startup with strong consistency. Pavlo had advised it to add SQL as a technical consultant, but the team did not follow his advice.
- PostgresML, which brought machine learning into PostgreSQL, did not take off either.
- The Hydra team, behind a DuckDB↔Postgres connector, dispersed.
- Apache Derby, a Java database dating to 1997, became “read only,” with development halted for lack of maintainers.
GPU databases had a difficult time too:
- Voltron Data, a startup supergroup with $110M in funding, did not manage to release its GPU-accelerated Theseus engine.
- HeavyDB, formerly OmniSci/MapD and a GPU database pioneer, quietly ceased operating independently as Nvidia absorbed its team.
Pavlo concludes that niche products such as Kinetica and Sqream remain peripheral, and GPU systems have not substantially displaced conventional CPU databases. Modern columnar CPU systems are already so fast that developer convenience and query optimiser intelligence matter more than raw optimisation. He nevertheless hints at new efforts to exploit GPUs fully in 2026: the race continues.
🏆 Larry Ellison at the top Pavlo describes a triumphant year for Oracle’s founder. At 81, Ellison became the world’s richest person, with wealth estimated around $393 billion. AI cloud deals drove Oracle’s valuation up, and its shares jumped 40% in one day in September 2025. In the review’s telling, that made Ellison No. 1 not just then but in inflation-adjusted history, ahead of Rockefeller. A couple of months later, Oracle’s valuation fell and he lost around $130 billion — like losing one salary for the rest of us, Pavlo jokes.
Oracle and Ellison also appeared in major deals: Oracle was a key partner in the effort to bring TikTok under US control and financed an attempted Warner Bros acquisition through the Ellison family’s Paramount holding company. Pavlo admits he was pleased that databases had made someone the richest person on Earth: finally, something positive for the field. He teases Ellison’s critics: having conquered enterprise databases, won regattas and built luxury spas for the technology elite, why should Larry not buy a news channel? Ellison does what he wants, supported by fans and his new wife; the rest does not matter to him.
#Database #Architecture #Management #DistributedSystems #Software #Engineering #Future