Open Source in 2025: The Year the Forks Worked

2025 was supposed to be the year open source fractured. License changes, consolidation, maintainer burnout—the doom narrative wrote itself.

Instead, something interesting happened. The forks worked. The community adapted. And open source ended the year stronger than it started.

Let’s talk about what actually happened, and what’s coming in 2026.

The forks worked

Remember when HashiCorp switched Terraform to the Business Source License? The open source community responded with OpenTofu, and a lot of people wondered if it would fizzle out.

It didn’t. OpenTofu got accepted into the CNCF as a Sandbox project in April. Fidelity announced they’re migrating from Terraform to OpenTofu. According to Spacelift, half of deployments are now on OpenTofu. The project has 3,900+ providers and 23,600+ modules.

The pattern held elsewhere. Redis went restrictive, and Valkey emerged under the Linux Foundation. Elastic’s license changes years ago spawned OpenSearch, now thriving at AWS and beyond.

The lesson: when companies change the rules, the community can fork and continue. It’s messy and takes time, but it works. The code survives.

PostgreSQL keeps eating everything

The 2025 Stack Overflow Developer Survey told a story everyone already knew: PostgreSQL won.

55.6% of developers now use Postgres—up nearly 7 percentage points from last year. That’s the largest single-year jump in the survey’s history. MySQL sits at 40.5%, and everything else is in the single digits.

Postgres has been “most wanted database” for four consecutive years. It’s the default choice for new projects, and increasingly the migration target for old ones. Instagram, Reddit, Spotify, NASA—the list of organizations running Postgres at scale keeps growing.

What makes this remarkable is how it happened. No VC hype cycle. No aggressive marketing. Just decades of solid engineering, a genuine community, and a simple pitch: it probably does what you need, and it’s free.

Local AI went mainstream

2025 was the year running AI locally stopped being a hobby project.

Ollama exploded—261% growth to over 105,000 GitHub stars. It turns running Llama, Mistral, or Gemma locally into a single command. No Python environments, no GPU configuration headaches. Just ollama run llama3.2 and you’re talking to a model on your own hardware.

The ecosystem around llama.cpp matured. LangChain passed 100k stars and became the default framework for building AI applications. Tools like Open WebUI gave local models a polished interface.

This matters beyond the obvious privacy benefits. When you can run capable models on a laptop, the economics of AI change. Not everything needs an API call to a cloud provider. The small models got good enough, and the tooling got easy enough, that local-first AI became practical.

Open protocols gained ground

2025 was a good year for the idea that protocols should be open and interoperable.

ActivityPub, the W3C standard powering Mastodon and the fediverse, got a major boost when Threads added a dedicated fediverse feed in June. With 350 million monthly active users, Threads is now the largest app running on ActivityPub—whether the fediverse wanted that or not. WordPress’s ActivityPub plugin hit version 7.x with full following capabilities, turning any WordPress site into a fediverse citizen.

Matrix, the open protocol for secure, decentralized communication, continued its march through government. Germany’s BundesMessenger is the world’s largest public-sector messaging deployment. France runs Tchap across all ministries with 80,000+ daily users. The German healthcare system is rolling out Matrix-based messaging to 74 million citizens. At the Matrix Conference 2025, representatives from 16 governments showed up.

The pattern is clear: when communication infrastructure matters—for sovereignty, security, or interoperability—open protocols win. Not because they’re ideologically pure, but because they’re practically superior. You can audit them, extend them, and switch providers without losing your data or your contacts.

Funding got serious

Germany’s Sovereign Tech Fund invested €23 million in 60 open source projects over two years. They’re funding systemd, PHP, Servo, curl, FreeBSD, FFmpeg—the unglamorous infrastructure everything else runs on.

GitHub commissioned a study on scaling this model EU-wide. The conversation shifted from “should governments fund open source” to “how do we make this work across Europe.”

The Open Source Pledge launched, asking companies to pay at least $2,000 per developer per year to maintainers. It’s not enough, but it’s a start—and it’s making the implicit dependency explicit.

The Linux Foundation announced the Agentic AI Foundation in December, with Anthropic, Block, and OpenAI as founding contributors. They’re taking stewardship of the Model Context Protocol (MCP), ensuring AI agent interoperability stays open.

The numbers don’t lie

From the 2025 State of Open Source Report:

• 96% of organizations increased or maintained their open source use
• 82% view open source as enabling innovation
• Open source AI/ML adoption jumped from 35% to 40%
• The open source services market is projected to hit $50 billion by 2026

97% of commercial codebases contain open source components. The average application includes over 900 of them. Open source isn’t an alternative anymore—it’s the default.

The challenges that remain

This isn’t all good news. Some real problems need solving:

The maintainer crisis is real. 60% of maintainers aren’t paid. About the same percentage have quit or considered quitting. The funding models are improving, but not fast enough. Eventually something critical breaks because a volunteer burned out.

“Open source AI” is still mostly marketing. Meta’s Llama and Google’s Gemma aren’t open source by the OSI’s definition—they’re open weights. You can run them, but you can’t reproduce them. The term is being diluted.

Consolidation continues. IBM closed the HashiCorp acquisition. More will follow. When infrastructure companies get acquired, community interests often take a back seat to enterprise sales.

These are real, but they’re not existential. The ecosystem has proven it can adapt.

What 2026 will bring

Here’s what I’m watching:

Local AI gets boring (in a good way). Running models locally will become as unremarkable as running a database locally. The tooling will mature, the models will shrink while staying capable, and “should we call an API or run it ourselves” will be a normal infrastructure decision.

More protocol-based apps. The fediverse proved the model works. ActivityPub, Matrix, and Nostr all have momentum. 2026 will see more apps building on open protocols instead of proprietary APIs—and more governments mandating them.

Government funding scales up. The Sovereign Tech Fund model will spread. More countries will recognize that open source infrastructure is critical infrastructure and fund it accordingly. The EU-wide fund is a question of when, not if.

State encryption becomes standard. OpenTofu shipped state encryption—something the Terraform community requested for five years. Features the community actually needs will ship faster in community-governed projects. This will accelerate adoption of forks.

PostgreSQL 18 and the AI workload. Postgres 18 brings async I/O and continues the project’s expansion into workloads that used to require specialized databases. The pgvector extension already made Postgres a viable vector database. Expect more AI-adjacent features.

WebAssembly finds its niche. WASM won’t replace containers, but it’ll become the default for plugins, edge computing, and sandboxed execution. The tooling is finally good enough.

The bottom line

Open source has never been more embedded in how software gets built. It’s also never been more contested—license changes, AI model confusion, and consolidation all challenge the ecosystem.

But 2025 showed the community can handle it. Forks work. Funding models are emerging. Open protocols are gaining ground. The philosophy of transparent, composable, community-governed software keeps winning.

2026 looks good. Let’s see what we build.

What are you most excited about for open source in 2026? I’m curious what trends people are watching.