Startup Blacksmith Reaches $1 Billion Valuation After $66 Million Round

Startup Blacksmith, which tests code, reached a valuation of $1 billion after a Series C round of $66 million led by Sequoia Capital, approximately two months after a Series B round of $45 million from Peak XV Partners at a valuation of $550 million. The total funding amounts to about $124.5 million.

More than 10,000 companies use the platform (in August there were about 5,000); among the clients are Granola, Handshake, Supabase, and LangChain. The number of checks increased from 5 million per week at the beginning of the year to 90 million, with annual recurring revenue growing 17 times to tens of millions of dollars. The largest clients spend more than $1 million a year. Blacksmith has about 40 employees in San Francisco and New York, with plans to increase the staff to 75 by the end of the year.

The company was founded in 2024 by Aditya Jayaprakash (CEO), Aayush Shah, and Aditya Maru – all of whom went through Y Combinator Winter 2024. Blacksmith offers continuous integration (CI): automated builds and tests before code merging and deployment. Most companies rent servers from AWS, Google Cloud, or Microsoft Azure for this, while Blacksmith buys its own machines – tens of thousands – and places them in data centers around the world.

According to Jayaprakash, developers used to split their time between writing code and waiting for checks. With AI tools like Cursor, Codex, and Claude Code, writing code has become "extremely easy," and validation has become the bottleneck. Anthropic reported that its own checks increased 25 times over six months, and after the launch of Claude Fable 5 in July, the volume of CI on Blacksmith increased by 40%. Since the beginning of 2026, the load has been growing by 5–10% per week.

Owning hardware provides control over performance but exposes the company to industry shortages of computing and memory. The main competitor is GitHub Actions from Microsoft; AI labs can also handle validation themselves. Blacksmith's response is the Codesmith agent, which diagnoses failed checks and fixes them. Jayaprakash insists that Blacksmith complements the labs: stronger models mean more code to validate.