Blacksmith Raises $45M to Validate AI-Written Code, Valuation Up ~10x in a Year

AI coding tools have made developers vastly more productive at producing code. The bottleneck is what happens next. However fast code gets written, it still has to be validated and tested before it ships — and that job falls to CI (continuous integration), the systems that automatically build and test every change to check it doesn’t break what already works. Right now, that gate can’t keep up.

One customer’s account captures it: “Our developers adopted Claude Code, PR volume increased 4x, and our CI infrastructure can’t keep up. We’re now having incidents almost every week that block engineers from merging code.” A PR, or pull request, is the standard unit in which a developer proposes merging their changes into the main codebase. AI raised the speed of writing code; the gate that verifies it did not move.

San Francisco-based Blacksmith attacks exactly that bottleneck with a purpose-built CI cloud, running the GitHub Actions workloads most developers rely on atop dedicated compute, caching, and storage tuned for CI.

The company announced a $45 million Series B led by Peak XV Partners, with existing investors Y Combinator and GV tripling down. The round values Blacksmith at $550 million — roughly 10x the $60 million valuation from its $10 million Series A less than a year ago. The company started out with a $3.5 million seed from GV and Y Combinator. It closed this round back in March, it says, but never got around to announcing it.

Its growth curve tracks AI coding adoption closely. Blacksmith says the release of Claude Opus 4.5 last December meaningfully changed how developers write software, and that CI jobs run on its platform have grown 5–10% week over week since the start of this year. More than 6,000 companies now run CI on Blacksmith, including Supabase, Clerk, Ashby, and Mercury. Per press reports, its customer count grew from 700 a year ago to more than 5,000; after crossing $10 million in annualized revenue with just 10 employees, revenue now sits in the “tens of millions” with a 30-person team.

Blacksmith was founded in 2024 by three University of Waterloo alumni. CEO Aditya Jayaprakash worked on search infrastructure and was a founding member of the ads team at Faire, while Aayush Shah and Aditya Maru were early systems engineers at Cockroach Labs, working on replication and disaster recovery respectively for CockroachDB. That hands-on infrastructure background is what led them to build a CI cloud.

Beyond CI, into the whole validation loop

The ambition extends past CI to a broader goal: helping developers validate and merge code faster. Its recently launched cloud coding agent, [code]smith, is the connective tissue — it builds features and fixes bugs, but also diagnoses and autofixes CI failures to keep PRs in a mergeable state. Next up is [code]smith QA, which will autonomously test changes before they merge.

Most of the money goes to compute. Blacksmith manages on the order of hundreds of thousands of cores today and plans to grow that footprint tenfold within months. “The main reason we raised this capital was to expand our compute footprint and get more aggressive about staying ahead of demand,” the company said.

Dev infrastructure becomes an AI-coding beneficiary

The downstream effects of AI coding are showing up across the stack. Backend platform Supabase raised $500 million at a $10.5 billion valuation on the back of vibe-coding adoption, and startups like Gitar have emerged to tackle “AI code validation” as the step after code generation. Entire, founded by GitHub’s former CEO, raised for an AI-agent collaboration platform. Capital that flooded into tools that write code is now flowing to the infrastructure that validates and runs it — and the question is whether Blacksmith can own that validation layer from CI through QA.