Overview of how Facebook's open-source projects, including Cassandra, GraphQL, the Open Compute initiative, PyTorch, and React, have changed computing
Klint Finley / Wired : Tweets: @wired Tweets: @wired : From Netflix to Uber to Walmart's website, many of the apps and services we use every day are built with technologies that Facebook developed and then shared with the world. http://www.wired.com/...
Context & Ripple Effects
This Wired retrospective caps a decade of Facebook giving away infrastructure rather than selling it. The Open Compute Project had already pulled hundreds of companies — HP, Foxconn, even Goldman Sachs — into jointly designed data center hardware, and Facebook went on to open-source its AI hardware designs while racing Google. The through-line: code Facebook gives away still dictates what everyone else builds on.
The counterpoint sits in the same coverage: a reported push to build an OS from scratch so Oculus and Portal can stop depending on Android. The giveaway strategy has limits — Facebook shares the layers where commoditization helps it and keeps control where product differentiation pays.
First-order effects
- Netflix, Uber, and Walmart run customer-facing services on Facebook-gifted stacks like React and GraphQL, meaning Facebook effectively sets the technology baseline for some of its biggest commercial rivals' web presence.
- Google finds itself competing against AI hardware designs Facebook publishes openly, turning proprietary R&D into a public benchmark.
Second-order effects
- Hardware vendors such as HP and Foxconn now co-design servers to Open Compute specifications alongside a bank like Goldman Sachs, shifting data center purchasing toward commodity designs Facebook helps standardize.
- Releasing internal tooling like the Horizon optimization tools on GitHub converts Facebook's Messenger-scale engineering into a recruiting magnet and de facto industry standard.
Third-order effects
- If the pattern holds, hyperscale platforms compete on operational scale rather than software ownership — while the reported from-scratch OS effort shows the same companies reasserting proprietary control at device and OS layers where openness would cost them leverage.
- Industry structure drifts toward a split: shared infrastructure (hardware specs, frameworks, ML tooling) that no one owns, atop tightly held differentiators (operating systems, recommendation systems) that decide who captures the value.
The trend: Major platforms increasingly give away core infrastructure to commoditize it, then reassert control at the layers — devices, operating systems — where differentiation still pays.