The Debate That Never Ends
The open source versus proprietary software debate has been running for four decades, and in 2025, it’s more nuanced — and more commercially important — than ever. The old “free versus paid” framing is obsolete. Today’s real questions are about governance models, licensing strategy, and who controls the infrastructure that runs the modern economy.
The Corporate Open Source Revolution
The biggest shift of the past decade isn’t that more companies use open source — they always did, whether they admitted it or not. It’s that open source has become the default development model for infrastructure software. Linux runs 96.3% of the top one million web servers and 100% of the top 500 supercomputers. Kubernetes won the container orchestration wars so decisively that nobody even debates it anymore. The $3 billion question — literally, what IBM paid for Red Hat in 2019 — was whether enterprise open source could be a serious business. The verdict is in: Red Hat generates over $5 billion in annual revenue selling support and enterprise versions of open source software.
But corporate involvement has changed open source. When a project’s core maintainers all work for the same company — think Google and Kubernetes, or Meta and React — is it really a community project, or is it a corporate product with a permissive license? The answer matters because corporate-controlled open source can change direction overnight if business priorities shift.
The Licensing Wars
Around 2018, a wave of open source companies changed their licenses in response to AWS and other cloud providers offering their software as managed services without contributing back. MongoDB switched from AGPL to the Server Side Public License (SSPL). Elastic did the same with Elasticsearch. Redis Labs (now just Redis) moved to a source-available license. HashiCorp switched Terraform from MPL to the Business Source License (BSL) in 2023.
The reaction was fierce. The Open Source Initiative (OSI) doesn’t recognize SSPL or BSL as open source licenses. Forks appeared: OpenSearch from Elasticsearch, OpenTofu from Terraform, Valkey from Redis. These forks are now serious projects backed by the Linux Foundation, AWS, and others.
As of 2025, the ecosystem has bifurcated. True community-governed open source projects thrive under foundations like CNCF, Apache, and Linux Foundation. VC-backed open source companies increasingly use restrictive licenses. The term “open source” itself is contested — what MongoDB sells isn’t open source by OSI’s definition, but it’s source-available and free for most use cases. Whether that matters depends on who you ask.
Linux Adoption: Dominant but Invisible
Linux’s dominance is so complete it’s become unremarkable. Android, the world’s most popular operating system, is Linux-based. ChromeOS is Linux-based. Every major cloud runs on Linux. The majority of embedded systems, from smart TVs to car infotainment, run Linux. Even Microsoft — the company whose former CEO called Linux “a cancer” — now runs more Linux than Windows on Azure, integrates a full Linux kernel into Windows via WSL 2, and is a top contributor to the Linux kernel.
The desktop is Linux’s last frontier, and even there the numbers are creeping up. Statcounter puts desktop Linux at about 4.5% market share in 2025, up from 2% in 2020. Steam Deck, Valve’s Linux-based gaming handheld, has sold millions of units and proved that Linux gaming is viable through Proton, Valve’s compatibility layer. It’s still a niche, but it’s a growing niche with real momentum.
The AI Conundrum
AI has introduced a new dimension to the open source debate. Meta’s LLaMA models are “open weights” — the trained model files are freely available — but not truly open source by OSI standards since the training data and methodology aren’t fully disclosed. The same goes for Mistral, Qwen, DeepSeek, and most “open source” AI models. You can download them, run them, and fine-tune them, but you can’t reproduce them from scratch.
The OSI has been working on an Open Source AI definition since 2022, expected to be finalized in 2025. The core question: can an AI model be considered open source if the training data isn’t available? The answer will shape AI regulation and corporate strategy for years.
Where We Are in 2025
Open source has won in infrastructure. The cloud runs on Linux, containers run on Kubernetes, and web servers run on NGINX or Apache. But the victory is qualified. Open source software is often built by underfunded volunteers while being monetized by billion-dollar corporations. The xz utils backdoor incident in 2024 — where a malicious actor nearly compromised SSH on most Linux systems through a trusted open source library maintained by a single burned-out volunteer — exposed the fragility of this arrangement.
The real question for 2025 isn’t whether open source is better than proprietary software. It’s how we sustain the open source ecosystem that modern civilization now depends on. Nobody has a great answer yet.