Blacksmith at $550M: CI is the new AI line item

Every dollar you save generating code with an agent, you spend somewhere else. Today's proof: Blacksmith, a CI cloud, raised at a $550 million valuation — roughly ten times its price a year ago — because AI coding tools produce more code than existing pipelines can validate. If your team adopted an agent this year and your CI bill jumped, that is not a billing error. That is the model.

What actually happened

TechCrunch reported that Blacksmith closed a $45 million Series B led by Peak XV Partners, with GV and Y Combinator returning. That values the company at $550 million, up from $60 million at its $10 million Series A less than a year ago, and brings total funding to $58.5 million.

The operating numbers are the interesting part, per the company's announcement:

• Customers grew from 700 to more than 5,000 in a year, including Mercury, Supabase, and Expensify
• ARR went from about $10 million to "tens of millions," with headcount going 10 to roughly 30
• Some of the largest customers spend over $1 million a year on CI

Blacksmith started as faster CI runners and has moved into the validation loop itself with codesmith, a cloud agent that diagnoses failed checks, autofixes them, and keeps pull requests green. CEO Aditya Jayaprakash put the thesis plainly: validating code is still a bottleneck, and a bigger one now because people are writing more of it.

Why the CI bottleneck matters for your business

Watch the direction of the money. A CI vendor with ten employees hit $10 million ARR, then grew customers 7x in twelve months without touching the code-generation market at all. It sells the check step. That is a signal about where AI-assisted development actually costs you.

The math is not subtle. An agent opens three PRs where a developer opened one. Each PR triggers the full suite. Each failed check triggers a retry loop the agent drives on its own. Your test minutes do not scale with features shipped — they scale with attempts, and agents attempt a lot. We have written before that review is the real bottleneck in AI-assisted development. Blacksmith's numbers are the invoice for that bottleneck.

Two practical moves for a small team:

Meter it. Put CI minutes on the same dashboard as your token spend. They are the same budget line now — cost per merged PR, not cost per run. If you cannot answer "what did this feature cost to validate," you are flying blind on the half of the bill that is growing.

Make the suite worth running. A test suite that is slow and flaky was an annoyance when humans ran it four times a day. When an agent runs it forty times, flakiness becomes a cash cost and a correctness cost — the agent starts "fixing" tests instead of code. Fast, deterministic, and trustworthy beats comprehensive. That is the same discipline behind testing your agents before production.

The vendor here is optional. The line item is not.

Sources: TechCrunch, Blacksmith via PR Newswire.