Blacksmith Raises $66M Series C to Expand Its CI Cloud

Ashby's developers had already tried bigger machines when their test jobs began waiting for somewhere to run. In Blacksmith's customer account, splitting tests across more runners exposed another problem: getting those runners assigned quickly enough. The engineering team needed capacity to arrive together, at the moment a code change asked for it.

Blacksmith is expanding the infrastructure behind that purchase with a $66M Series C announced October 21, 2026. Sequoia Capital led the financing at a $1B valuation, joined by Premji Invest and existing investors Peak XV Partners, GV and Y Combinator. The San Francisco company puts total funding at $125M and plans to use the new capital to expand its team and compute footprint.

The financing follows a change in the work around software development. AI can produce more proposed changes, but engineering organizations still need machines to build those changes, run tests and return evidence a reviewer can use. Blacksmith makes a business of supplying that infrastructure while customers keep working inside GitHub Actions.

Blacksmith buys capacity so customers can stop managing the fleet

Continuous integration, or CI, runs builds and tests against changes to software. Blacksmith's product documentation describes a managed service that replaces the runner executing a GitHub Actions job. The workflow remains recognizable to the engineering team while the supplier takes responsibility for the compute underneath it. Caches, logs and test analytics sit around that execution layer.

That division of labor gives an infrastructure buyer several things to evaluate at once. A lower charge per minute has limited usefulness when the job spends its time waiting to start. Faster execution also depends on getting dependencies and build artifacts close to the machine doing the work. Blacksmith's homepage describes high-performance CPUs, colocated caches and persistent Docker layers as parts of its approach, rather than treating processor speed as the whole product.

Ashby's experience illustrates why those parts belong in the same buying decision. The vendor-hosted case study describes tests divided into 31 parallel shards and demand rising from 0 to 3,500 vCPUs in seconds. Its reported savings and deployment improvements belong to that particular customer implementation. The useful operating detail is the coordination: parallel tests need available machines together, or the organization can trade one long task for a collection of shorter tasks waiting in a queue.

For a team adopting the service, retaining GitHub Actions also reduces the amount of process change attached to a trial. Engineers can examine the runner's contribution without first rebuilding the way the organization expresses its workflows. The supplier still has to earn trust through the behavior of the workloads it receives.

The founders bring distributed-systems experience to an uneven workload

Blacksmith was founded in 2024 by Aditya Jayaprakash, Aayush Shah and Aditya Maru. Y Combinator's active-founder profiles identify Jayaprakash as cofounder and CEO, Shah as cofounder and CTO, and Maru as cofounder and co-CTO. The three University of Waterloo graduates brought engineering experience from Faire and Cockroach Labs into a company serving other engineering teams.

Their earlier employers matter as technical background, rather than as borrowed prestige. Building and operating distributed software exposes engineers to workloads that grow unevenly and failures whose consequences travel beyond the first machine. Blacksmith has chosen to sell a service where the customer's convenience depends on the supplier continuously doing that operational work. Capacity planning remains inside the purchase, even when the customer no longer handles it directly.

The funding history also deserves a calendar read more careful than a sequence of headlines. Blacksmith's $45M Series B announcement appeared on August 12, 2026, but Jayaprakash wrote that the round was raised in March. That financing, led by Peak XV Partners, valued the company at $550M. Its $10M Series A, led by GV, was announced in September 2025.

The Series C therefore follows capital raised earlier in the year, even though the public announcements arrived closer together. Sequoia is now backing additional infrastructure alongside a returning investor group already familiar with the company's development. The $1B valuation describes the financing; the operating story still runs through the customers sending work to the fleet.

More generated code changes the economics of checking it

Blacksmith reports that weekly CI jobs grew from 5M in January to more than 90M, while annual recurring revenue increased 17x. More than 10,000 companies use its platform, according to the company. These are supplier-reported measures, and the announcement leaves absolute ARR and unit economics undisclosed.

Job volume is useful evidence of the service's workload, but it cannot explain the business on its own. Different jobs consume different resources and run for different lengths of time. An infrastructure company must provide enough capacity to handle demand while maintaining a service customers find worth buying. Reading job growth alongside revenue growth keeps those separate questions visible without inventing a margin figure.

There is independent first-party evidence that AI adoption can strain the checking process. In a September engineering account, Anthropic described a 25x increase in CI jobs over 6 months and the redesign of its test-selection service. That account concerns Anthropic's own infrastructure; it establishes neither a Blacksmith customer relationship nor a universal growth rate for the industry.

For software leaders, the connection is practical. Generating another proposed change adds work downstream: selecting checks, executing them, interpreting failures and deciding whether the change can merge. Spending on coding assistance can therefore create another infrastructure purchase. DevCuration's coverage of Daytona's agent cloud and Asymptote Labs' agent-control infrastructure traces adjacent purchases around the same expanding development workflow. The commercial opportunity for Blacksmith rests in making that additional work manageable inside the tools a customer already uses.

The new capital reaches the data center

Blacksmith says it manages just under 10 MW of compute across US and European data centers. Its Series C announcement sets a target of multiplying that footprint by 5 in 2027, alongside team expansion. This is a management plan for future capacity, rather than a description of equipment already available to customers.

The purchase has a physical timetable. A team can change a workflow configuration quickly; the supplier has to prepare the infrastructure that makes the change useful. Blacksmith's investment therefore joins software product development with the less visible work of ensuring machines are ready when engineering demand arrives.

Its documentation also shows an expanding validation ambition, with the early [code]smith agent and private-beta Proof alongside established CI products. Those maturity labels matter to a buyer considering how much of a development process to entrust to the company. Running a test supplies evidence about the checks performed; engineering organizations retain responsibility for deciding which checks their software requires.

For Jayaprakash, Shah and Maru, the expanding fleet makes that relationship more consequential with each new customer. Their service sits at the point where an engineer or agent has finished proposing a change and needs the organization to examine it. The next burst of builds will arrive on the customer's schedule, with the expectation that Blacksmith has already made room.