A Melbourne company runs the testing for OpenAI and Anthropic. Its latest accounts show it spent more than $50M in a year to bring in $43.2M, and that it quietly raised $21M in January.
What it does
Every time OpenAI, Anthropic, Cursor, Uber, Airbnb or Canva changes their software, that change gets tested before it ships. Buildkite runs those tests, and it runs them on the customer’s own machines rather than someone else’s cloud. It has been doing this since 2013.
The numbers
The FY25 accounts lodged with ASIC, reported by Capital Brief, show revenue up 16% to $43.2M. Staff cost $39.1M and hosting $11.3M. The net loss more than doubled to $28.4M. Cash fell from $33.7M to $10.1M. The $21M raise, via a convertible note in January 2026, only surfaced once the accounts were lodged.
The turn
In a May job ad, Buildkite said workloads had doubled in three months, with customers wanting 5 to 10 times more capacity across 2026. The reason is AI coding agents. They write far more code than people do, and every line of it needs testing.
My read
These shifts take years, not months, and years is plenty of time to change what you sell. Buildkite has been around since 2013 and has already outlived a few things that were meant to replace it. If agents keep writing more code, the testing layer gets busier, not quieter.