GitHub Copilot had 20 million users and first-mover advantage. It did not matter. After Claude Code and OpenAI Codex launched in mid-2025, Copilot daily installs peaked and reversed within six months while competitors crossed 100,000 combined installs. The early leader lost ground faster than any laggard. That is the new baseline for software competition.
The author analyzed 374 quarterly net dollar retention observations across 25 public software companies. From 2022 to 2025, NDR declined gradually, from 125% to 112%, with no single shock tied to ChatGPT or enterprise Copilot adoption. Then 2026 arrived. The bottom quartile fell from 106% to 101% NDR in one quarter, statistically significant at p less than 0.0001 with R-squared of 0.74. Zoom sits at 98%. Asana at 96%. Bill.com at 94%. The bottom quartile is now in contraction. Asana growing 9% annually with 96% NDR requires 13-plus percent new customer acquisition just to hold flat revenue.
The analysis is worth reading in full because it separates three distinct threats hitting the bottom quartile simultaneously: macro pressure on SMBs, where bankruptcies hit a 15-year high in 2025; commoditization from near-free competitors like Teams and Google Meet; and AI agents replicating simple task workflows. Tunguz does not predict uniform collapse. He maps which companies face which blades and why products with narrow feature surfaces are most exposed. The Copilot data is the opening argument. The NDR breakdown is the proof.
[READ ORIGINAL →]