OPEN SOURCE SOFTWARE

Open source software is software whose source code is publicly available for anyone to use, modify, and distribute — a development model that emerged in the 1980s and 1990s and now underpins the vast majority of web servers, cloud infrastructure, programming languages, databases, and AI frameworks.

The Business Model

The economics of open source are counterintuitive: how do you build a business on something you give away? The answer is that open source is a distribution strategy, not a revenue model. Companies have converged on three primary monetization paths.

Hosted services. Cloud providers (and the projects themselves) run open source software as a managed service. Users pay for convenience, reliability, and scale — not the software. This is the software-as-a-service logic applied to open source: it is how MongoDB, Elastic, and Redis Labs generate revenue from projects they open-sourced. It also explains why AWS, Google Cloud, and Azure profit enormously from open source infrastructure they did not create.

Enterprise support and features. Red Hat built a billion-dollar business — acquired by IBM for $34 billion — selling support contracts and enterprise tooling around Linux. The open source project is the proof of quality and the distribution channel; the enterprise offering is the product.

Dual licensing. Some projects offer an open source license for developers and a commercial license for businesses that want to embed the software in proprietary products without open-sourcing their own code. MySQL pioneered this model; it remains common in developer tools and databases.

The AI Dimension

Open source AI models — Llama, Mistral, Falcon — are challenging the commercial model of frontier AI labs. When capable models are freely available, the economic moat of proprietary model weights erodes. The competitive response from commercial labs has been to improve faster than open source can keep up, build application-layer lock-in, and emphasize safety and reliability advantages.

Whether open source AI eventually commoditizes frontier capabilities — as Linux commoditized server operating systems — is the central structural question of the AI industry right now. The analogy is imperfect: training frontier models requires compute at a scale that individual contributors cannot replicate, which gives commercial labs a structural advantage that the Linux Foundation never had to overcome.

For Builders

Open source has dramatically lowered the cost of building software. Libraries, frameworks, and models that would have required years and millions of dollars to build are freely available. The constraint has shifted from access to capability to judgment: which open source components to trust, how to contribute back, and when to build proprietary foundations versus open ones.

The build-versus-open decision has real strategic weight. Building on open source accelerates development but can limit differentiation. Releasing proprietary software as open source can accelerate adoption but surrenders pricing power — unless a monetizable layer sits above it.

The Sustainability Problem

Open source creates a structural imbalance between value creation and value capture. A library maintained by two volunteers can underpin billions of dollars of commercial software infrastructure. The Log4Shell vulnerability in 2021 exposed this sharply: a critical flaw in a small, under-resourced project threatened the entire Java ecosystem for weeks.

The solutions that have emerged — foundation models like the Apache Software Foundation, corporate sponsorship, dual-licensing — all have weaknesses. The open source projects most likely to remain well-maintained are those with a clear commercial stakeholder whose business depends on the project's health, or those with a contributor community large and distributed enough that no single point of failure exists. Everything in between is fragile.

Open Questions

• Can open source AI models close the capability gap with commercial frontier labs, or does training compute create a permanent structural advantage for well-capitalized players?
• Does dual-licensing work long-term, or do competitors fork projects before restrictions kick in — as happened with Elasticsearch and OpenSearch?
• Who is responsible for the security of open source infrastructure that has become critical to global systems but receives no proportional funding?
• As AI-assisted coding lowers contribution costs, does the supply of open source maintainers increase — or does it increase low-quality contributions while maintainer burnout stays constant?

Part of the knowledge graph at The Best Blog Ever — reference definitions for ideas that matter.