Code-Testing Startup Blacksmith Raises $66 Million at a $1 Billion Valuation

Blacksmith, a startup that provides computing infrastructure for building and testing software, announced a $66 million funding round Wednesday led by Sequoia Capital. The deal values the company at $1 billion and finances a business positioned around a practical consequence of AI-assisted coding: producing more code also creates more work to verify before release.

The Series C includes Premji Invest and returning investors Peak XV Partners, GV and Y Combinator. Blacksmith said total funding has reached $125 million and that the new capital will support hiring and a larger computing footprint. The financing is an investment in a privately held company, rather than a public stock offering available to ordinary brokerage customers.

The work after an AI writes code

Generating a software change is only one part of delivering a functioning product. A development team still needs to assemble the application, run automated tests and investigate failures before accepting the change. Continuous integration, commonly shortened to CI, automates much of that work whenever developers submit changes to a shared project.

The business opportunity is straightforward. If developers and coding agents submit more changes, the computers running those checks can become a bottleneck. Faster code production may then increase waiting time instead of accelerating delivery. Blacksmith sells the infrastructure for executing that validation workload; it does not establish that every submitted change is correct or that automated tests catch every defect.

Blacksmith reports that weekly CI jobs on its platform increased from 5 million in January to more than 90 million, while annual recurring revenue grew seventeenfold. It also says more than 10,000 companies use the service. These are company-reported operating figures, not independently audited financial results, and the announcement does not disclose the absolute revenue base behind the growth multiple.

A valuation is different from revenue

A $1 billion valuation describes the pricing of an ownership stake in the financing. It is not annual sales, cash in the bank or profit. Likewise, the $66 million round should not be confused with the company's total funding. Keeping those measures separate is especially useful when rapid usage growth makes a young infrastructure business appear larger than its disclosed financial record can establish.

Annual recurring revenue is also a different measure from revenue recognized in financial statements. Without the underlying definition, customer contracts and revenue disclosures, the seventeenfold figure cannot support a reliable calculation of a valuation-to-sales multiple. The release provides evidence of investor support and reported expansion, while leaving important profitability and customer-retention questions unanswered.

Our analysis of the financing demands behind the AI boom examines the same distinction at a broader level. Capital can pay for capacity before the operating business produces enough cash to finance expansion itself. The commercial test is whether customers continue buying the service at prices that cover its full costs.

Faster tests still have to earn their cost

Blacksmith's announcement advertises faster jobs and lower average costs, but those claims should be evaluated against a customer's actual workload. A short test with little computing demand and a large application build can have very different constraints. Moving execution to faster infrastructure will not repair a poorly designed test or remove an external service that slows a pipeline.

For buyers, a useful comparison includes waiting time, execution time, failure investigation and the engineering effort required to operate alternatives. A lower computing bill is only part of the calculation. Reliability matters because interrupted checks can delay a release, while weak test coverage can allow defects through even when every automated job finishes successfully.

Security remains part of the purchase

The company's security documentation says it integrates with GitHub Actions and executes jobs in isolated Firecracker micro virtual machines that are destroyed after completion. That describes an important operational boundary, rather than a guarantee against every possible security incident. Build systems can handle sensitive code, dependencies and credentials, making access controls and the treatment of stored artifacts relevant purchasing questions.

For Blacksmith, the funding creates room to meet a growing workload while competing on speed, price and operational trust. The figures to watch next are repeat customer spending, service reliability and the relationship between computing costs and revenue. More AI-generated code expands the potential workload; retaining that workload profitably is the separate business challenge.