Blacksmith raises $45M to aid AI code validation as agentic development grows
SiliconANGLE Kyt Dotson ● Covered by 2 sources
Blacksmith just raised $45M to test AI-written code in the cloud. More agents are shipping code, and the boring validation step is becoming the bottleneck.
Based on reporting by SiliconANGLE, Kyt Dotson — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Blacksmith Software Inc. has raised $45 million to push more of software testing into the cloud, right where its customers are already writing and validating code. The Series B was led by Peak XV Partners, with Y Combinator and GV joining in. The round values the company at $550 million.
The timing makes sense. More developers are leaning on agentic tools like Claude Code and Codex, which can spit out a lot of code fast. That creates a new problem: the code still has to be built, tested and checked before anyone trusts it. Blacksmith is betting that the dull part of development is now the critical part.
Its newer product, codesmith, is aimed at that gap. Blacksmith describes it as a cloud coding agent that works inside the validation loop, spots integration failures, fixes them and heals submissions to the codebase. In practice, the company says teams can keep working in parallel, send changes in together, and let Blacksmith test each submission as it arrives. Passes get committed. Fails get bounced back with notes.
The money will go into compute infrastructure. Blacksmith says it already manages hundreds of thousands of cores and plans to increase that by an order of magnitude in the coming months. Demand is already moving: the company says jobs running on its service are up between 5% and 10% week over week since the beginning of the year.
The customer list has grown too. Blacksmith says more than 6,000 companies now use its service, including Supabase Inc., Clerk Inc., Ashby Inc. and Mercury Systems Inc. That is a big jump from about 800 customers when it announced its Series A last September. In other words, the company is riding a very specific wave: not code generation, but the messy, unglamorous work of proving the code actually works.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI coding everyone keeps pretending is optional: checking the machine’s homework. The agents get the headlines, but the validation layer is where the actual business lives, because broken software still has to be caught before it embarrasses someone in production. Very glamorous, very necessary, very 2026.
Read more about this at: SiliconANGLE