Claude Code reached $2.5 billion in annualised recurring revenue as of February 2026. The previous milestone — $1 billion ARR — was crossed in Q4 2025, roughly six months after public launch. The jump from $1B to $2.5B took about three months.
For context: it took GitHub Copilot years to reach comparable revenue. Claude Code did it in under 18 months.
At the Code with Claude conference in May 2026, Anthropic CEO Dario Amodei said the company had modelled 10x annual growth and saw 80x, annualised, in Q1 2026. That’s not a metric anyone predicted.
What the number actually means
$2.5B ARR at the team and enterprise tier means development organisations are paying for these tools at scale, not experimenting with free trials. It means AI coding assistants are now a recurring line item in engineering budgets, like cloud hosting or version control.
It also means the development teams building your software are almost certainly using one of these tools. Claude Code at 80.9% on SWE-bench Verified leads the benchmark leaderboard in mid-2026. Cursor reached $2B ARR around the same time. The market has three or four dominant tools — Claude Code, Cursor, GitHub Copilot, and now Devin Desktop — and they’re all at or approaching the $1B+ ARR threshold.
What it changes for commissioning conversations
The practical implication for founders and product leads is not that AI coding tools exist — you already knew that — but that the teams you hire are now expected to be proficient with them. A development team in 2026 that isn’t using these tools is working at a competitive disadvantage on speed and cost.
This also means “AI-assisted development” is no longer a premium service differentiator. It’s a baseline expectation. The question worth asking your development partner isn’t “do you use AI tools?” — it’s “how do you use them, what do you review manually, and how does that affect the timeline and price we’re discussing?”
The explosive growth in AI coding tool revenue happened because the tools genuinely accelerate delivery when used well. The risk is using them poorly — at volume and without oversight — which is where the security and quality debt accumulates.
Our AI-assisted development approach is built around the second question, not the first. If you want to understand what that looks like in practice, start a project conversation.