Blacksmith Series B: $45M as AI code floods CI
Blacksmith raised $45M to run CI faster. The customer quote in the announcement is the real story: adopt an AI coding agent, watch pull request volume go up 4x, watch CI fall over.
The Blacksmith Series B, announced 12 August 2026, is $45 million led by Peak XV Partners at a $550 million valuation, with Y Combinator and GV following on. Blacksmith runs continuous integration jobs, which is not a glamorous business, and the reason it just got funded is a bottleneck most teams meet about three weeks after they start using an AI coding agent seriously: the code arrives faster than anything can check it.
Blacksmith Series B by the numbers
From Blacksmith's announcement and TechCrunch's coverage:
$45 million Series B, led by Peak XV Partners, at a $550 million valuation.
More than 6,000 companies on the platform, up from roughly 800 at the Series A last September.
CI jobs on the platform growing 5 to 10 percent week over week since the start of 2026.
Hundreds of thousands of compute cores under management, with plans to grow that by an order of magnitude.
Most of the money goes to compute. That is the tell. This is not a round about a clever product idea, it is a round about buying enough machines to keep up.
The quote that explains the round
Our developers adopted Claude Code, PR volume increased 4x, and our CI infrastructure can't keep up.
That is a customer, quoted in the company's own announcement. It describes a failure mode that is easy to miss when you are measuring the wrong thing. Teams adopting AI coding agents track how much code gets written. The constraint moves somewhere else: to the pipeline that has to build, lint, test and review everything the agent produced.
If a test suite takes eleven minutes and pull request volume quadruples, the queue does not grow by four. It grows until developers stop waiting and start merging on faith, which is the actual danger. Slow CI does not usually block bad code. It teaches people to skip the check.
What codesmith is doing
Blacksmith also launched codesmith, a cloud agent that sits inside the validation loop rather than the authoring loop. It diagnoses CI failures, fixes them, and keeps pull requests green, with a QA product planned that tests changes before merge.
The design choice is interesting on its own. Almost every AI coding product to date has aimed at the writing step. Pointing an agent at the failures instead is an acknowledgment that the writing step is no longer where the time goes. If you already run an AI coding agent in CI, this is the same idea being sold as infrastructure.
It also raises the obvious question. An agent that autofixes failing checks is an agent that can make red things green without anyone reading the diff. Green is a signal about the tests, not about the code, and it means less when a machine was optimising for the colour. That does not make it a bad tool, but it moves more weight onto reviewing AI generated code before you ship it and onto whether your tests actually assert anything.
Why this matters if you are not running a platform team
Solo builders and small teams do not have a CI capacity problem at 6,000-company scale. They have the same problem in miniature, and usually without noticing:
Volume goes up. An agent producing five changes an afternoon is normal now.
Verification stays manual. One person still reads everything, or nobody does.
The check quietly becomes optional, because it is the slow part.
The fix at small scale is not more compute. It is having tests worth running, which is a good argument for spending some of the agent's speed on using AI to write tests rather than only on features. A funding round of this size is essentially a market-wide confirmation that verification, not generation, is where the next few years of AI coding tools will be fought. That is the part worth filing away, and it is a better reason to follow funding announcements than the valuations are, which is roughly the argument for a deliberate habit for tracking AI releases.
FAQ
How much did Blacksmith raise?
$45 million in a Series B led by Peak XV Partners, at a $550 million valuation, announced on 12 August 2026. Y Combinator and GV also participated.
What does Blacksmith actually do?
It runs continuous integration jobs on its own compute, aiming to make builds and test runs faster than they are on general-purpose CI runners.
What is codesmith?
A cloud coding agent from the same company that works inside the validation loop: diagnosing failing CI checks, fixing them, and keeping pull requests passing. A QA version that tests changes before merge is planned.
Should an agent be allowed to fix its own failing tests?
Only if a human still reads the diff. An agent optimising for a green pipeline can make checks pass without making the code correct, which is a worse outcome than a visible failure.
How did this land?
About the author
Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.
