The database that lets everyone take its code for free just won its fourth "Database of the Year" award. Here's why that's not an accident — it's a masterclass in platform economics.
Walk through any tech conference in Beijing or Shenzhen, and you'll hear a familiar refrain: Chinese database companies have been "freeriding" on PostgreSQL for decades. openGauss — Huawei's flagship database — was forked from PostgreSQL 9.2.4. Alibaba's PolarDB boasts 100% PostgreSQL compatibility. KingbaseES, the go-to database for Chinese government systems, traces its lineage to PostgreSQL 9.6. The list goes on: HanGao, Hailiang, Youxuan — all building commercial, closed-source products on a BSD-licensed foundation they never paid a cent for.
The term used in Chinese tech circles is "白嫖" (báipiáo) — literally "white嫖", slang for getting something valuable without paying. And legally speaking, they're right. PostgreSQL's license — a BSD-style agreement unchanged since 1996 — explicitly permits exactly this: use it, modify it, sell it. Just keep the copyright notice.
So here's the billion-dollar question: When MongoDB, Elastic, and Redis all fought back against cloud vendors by tightening their licenses, why did PostgreSQL — the most freerided database on Earth — do absolutely nothing?
The answer isn't passivity. It's one of the most elegantly rational strategies in open-source history.
Let's start with the structural reality. PostgreSQL isn't a company. It has no shareholders. No board. No CEO. No one with the authority to bang a gavel and declare a license change.
The PostgreSQL Global Development Group (PGDG) is closer to a academic research collective than a business entity. It has no legal standing whatsoever — it's an abstract term describing the global community of developers who contribute to the project. Actual governance is handled by a seven-person Core Team: long-time community members appointed internally, never elected, with no corporate mandate.
This structure is deliberately license-change-proof. Here's what would actually have to happen for PG to adopt a restrictive license like SSPL:
First, a proposal would need to reach consensus on the global mailing list — where 90% of license change proposals die immediately. Why? Because PG's committers come from dozens of competing organizations: EnterpriseDB, Microsoft (which acquired Citus in 2021), Amazon, Google, Red Hat, Postgres Pro (Russia), SRA OSS (Japan), HighGo (China), and scores of independent contributors. No single employer accounts for even close to 50% of committers.
Second, even if consensus somehow materialized, the legal machinery is impossible. PG's Contributor License Agreement (CLA) doesn't assign copyright to any single entity — it grants PGDG the right to relicense, but only within the bounds of the existing permissive framework. Changing to SSPL would require tracing 30 years of contributions from thousands of developers across dozens of jurisdictions. The legal costs alone would run into the tens of millions.
Compare this to MongoDB Inc., Elastic NV, and Redis Ltd. — each a single commercial entity where key contributors signed CLAs assigning full copyright to the company. A board vote. A press release. Done.
PostgreSQL's governance isn't broken. It's a fortress. And the moat was dug on purpose.
If the governance argument explains why PG can't change, the economic argument explains why it shouldn't — even if it could.
Think of PostgreSQL as a three-sided platform:
The conventional complaint is that Side C exploits the platform without paying. But platform economics tells a different story: Side C is a positive feedback amplifier, not a parasite.
Every time Alibaba launches a PolarDB-compatible service, it expands the PostgreSQL ecosystem. Every time openGauss wins a government contract, it trains thousands of developers who now know the PostgreSQL wire protocol, SQL dialect, and operational patterns. Every cloud provider offering Postgres-as-a-Service reduces the friction for Side B adoption.
The numbers tell the story. PostgreSQL has been named DB-Engines Database of the Year four times (2017, 2018, 2020, 2023). Stack Overflow's 2024 developer survey placed it at 48.7% usage — number one among all databases for two consecutive years. Over 83,000 companies started using PostgreSQL in 2025. It's the default database for AI and machine learning workloads, powering everything from vector search to training data pipelines.
Now run the counterfactual. What if PG adopted SSPL tomorrow?
The revenue side: six major clouds (AWS, GCP, Azure, Alibaba, Tencent, Huawei) operating PG-as-a-Service at an estimated global market of $8-15 billion annually. At MongoDB-style royalty rates, that's $120-450 million per year in potential licensing fees.
The loss side: contributor activity drops 30% (consistent with what happened to Elastic post-2021). At least 5% of self-hosted users flee to MySQL or MariaDB to avoid license uncertainty. And 80% of Chinese vendors go the "license-independent fork" route — exactly like AWS did with OpenSearch. The net present value over five years? Almost certainly negative.
PostgreSQL's BSD license is what economists call a commitment device — a credible promise that the platform will never turn extractive. That promise has been kept for 30 years, and it's the single most valuable asset PGDG owns.
If the theory isn't convincing, the experiments have already been run — at enormous cost.
MongoDB Inc. switched from AGPLv3 to the Server Side Public License (SSPL), targeting cloud vendors who offered MongoDB as a managed service without contributing back. The Open Source Initiative formally rejected SSPL as not meeting the Open Source Definition. Debian, Fedora, Red Hat Enterprise Linux, and even Homebrew removed MongoDB from their package repositories. Revenue growth guidance dropped from +30% to 15-18%. And while MongoDB's stock eventually recovered, its DB-Engines popularity score peaked in early 2021 and has been declining ever since. A 2024 Redmonk study found "no clear link" between open-source license changes and increased company value.
Elastic moved Elasticsearch and Kibana from Apache 2.0 to a dual license of SSPL + Elastic License. The result? AWS forked both projects into OpenSearch within three months. OpenSearch attracted 496 contributors and over 100 million downloads in its first year. By September 2024, Elastic effectively admitted defeat — adding AGPLv3 as a third option, a move widely interpreted as a public rollback. CTO Shay Banon claimed the original change "worked" because AWS is now "fully invested" in OpenSearch — a remarkable reframing of "we created a permanent, institutionally-backed competitor."
Redis Ltd. switched from BSD-3-Clause to RSALv2 + SSPLv1. Within days, AWS, Google Cloud, Oracle, and Ericsson forked the last BSD-licensed version (7.2.4), named it Valkey, and donated it to the Linux Foundation. Within months, AWS made Valkey the default engine for ElastiCache and MemoryDB. Google Cloud followed suit in Memorystore. Independent benchmarks showed Valkey delivering 37% higher throughput than Redis 8.0. Today, Valkey has nearly 50 contributing companies and the project is developing features that increasingly diverge from upstream Redis. The fork didn't just survive — it's winning.
In every case, the pattern is identical: a single company tightens its license to capture cloud revenue → the cloud ecosystem forks the last open version → the fork becomes a permanent, foundation-backed competitor → the original project loses more than it gains.
Nowhere is PG's permissive strategy more visible than in China. The Chinese distributed database market reached $420 million in the first half of 2025 alone. And PostgreSQL DNA runs through virtually every major player:
Every single one of these is legally built on PostgreSQL's BSD license. And yet here's the uncomfortable truth the original Toutiao article identified: not one of them has achieved genuine international relevance. openGauss doesn't compete with Oracle in Frankfurt. PolarDB doesn't challenge AWS RDS in Virginia. OceanBase doesn't win contracts in London.
Why? Because PostgreSQL compatibility is table stakes, not a differentiator. When every competitor can legally embed the same kernel, the competitive moat shifts elsewhere: service quality, compliance certifications, SLA guarantees, ISV ecosystems, developer tooling, global support infrastructure. These are hard things. They require investment measured in decades, not quarters.
PostgreSQL's refusal to change its license isn't naivety, weakness, or institutional paralysis. It's a Nash equilibrium — a stable state where no player can improve their position by changing strategy.
PGDG looked at the same data MongoDB, Elastic, and Redis looked at. It saw cloud vendors monetizing open-source code without contributing proportionally. It saw Chinese companies building nine-figure businesses on PG's kernel. It saw the same "freeriding" that drove its peers to reach for restrictive licenses.
And it made a different calculation: the BSD license isn't a bug. It's the engine.
That permissive license has attracted contributors who would never touch a GPL project. It's enabled cloud providers to build managed services that make PostgreSQL the path of least resistance for enterprise adoption. It's allowed Chinese database companies to train an entire generation of developers on the PostgreSQL protocol — developers who will reach for PostgreSQL-compatible tools for the rest of their careers.
The framing of "freeriding" misses the point entirely. PGDG isn't a victim. It's a platform operator in a repeated game who understood that the long-term equilibrium strategy is zero rent extraction. Every dollar of license fees forgone is an investment in ecosystem growth that compounds at a rate no royalty check can match.
So the question worth asking isn't "Why won't PostgreSQL fight back?" It's the one the original Chinese article posed: Why, after 30 years of unlimited access to one of the world's best codebases, has no PostgreSQL derivative — Chinese or otherwise — built a truly irreplaceable business?
The answer to that question has nothing to do with licenses. And everything to do with what it actually takes to win.
