Google’s Secret Plan to Replace C++
What if the world’s largest C and C plus plus codebases could evolve without stopping the assembly line, without rewriting millions of lines, and without surrendering performance? For decades the status quo bent under its own weight. Backwards compatibility preserved every rough edge. Build times stretched. Tooling groaned. Memory safety incidents kept appearing. Production systems in browsers, operating systems, finance, and telecom met a hard truth at scale: when the language cannot retire old debt, the debt taxes every change. The question was not whether C plus plus could keep going. It was whether teams could keep paying the bill forever.
Before Carbon, teams lived inside constraints that showed up on every dashboard. Cold builds took tens of minutes on large monorepos, and incremental compiles still spiked into minutes when templates exploded. Peak memory for builds ballooned into tens of gigabytes when metaprogramming met header duplication. Link steps remained slow and opaque. Debug configurations fought a thicket of undefined behavior that sanitizers could catch only after the fact. Reliability suffered in ways that matter to executives as much as to engineers: chromium security teams have reported that roughly seventy percent of serious bugs trace to memory safety, and Android security teams showed multi-year charts where shifting to memory safe code steadily dropped the share of memory vulnerabilities from over three quarters to under one quarter. The lesson was clear. At scale, time, memory, and reliability were not abstract properties. They were invoices.