Software needed to run every Google internet service spans 2B lines of source code, all in a single repository available to all 25K engineers
Google Is 2 Billion Lines of Code—And It's All in One Place — How big is Google? We can answer that question in terms of revenue or stock price …
Context & Ripple Effects
Wired's 2015 look inside Google's monorepo — two billion lines serving every internet service, visible to all 25,000 engineers — was framed as an engineering-culture story about tooling and scale. What it actually documented was the creation of a single, uniform surface over all of Google's code.
That surface is what makes the later numbers legible. Pichai's quarter-of-new-code-is-AI disclosure in 2024, the jump to 75% by spring 2026, and the strike team Sergey Brin tasked with catching up on coding agents all presuppose a codebase where generated changes can be reviewed, tested, and accepted against everything else. The open-source practices site Google published two years after this piece showed how deliberately it managed that boundary between internal and external code.
First-order effects
- All 25,000 engineers work against one dependency graph, so any change to shared code is immediately visible and reviewable across Search, ads, and consumer products alike.
Second-order effects
- A centralized repository is the precondition for Google's AI-codegen ramp: models generating most new code need a consistent corpus and a single review pipeline, which is exactly why Brin pushed DeepMind toward coding agents on top of it.
Third-order effects
- If the pattern holds, the repository itself becomes strategic infrastructure — a moat that compounds as more code is machine-generated against it — shifting competitive advantage from headcount to whoever owns the largest reviewed code corpus.
The trend: Engineering infrastructure built for human collaboration at scale is being repurposed as the substrate on which AI writes and reviews code, turning monolithic codebases into AI-era assets.