Blacksmith

Ask most engineers where their day disappears and, sooner or later, someone will name the culprit: the pipeline. The commit is pushed, the checks spin up, and then everyone waits. Blacksmith, a San Francisco startup founded in 2024, was built around a blunt observation - that continuous integration is often the slowest and most expensive part of shipping software, and that it does not have to be.

Blacksmith runs GitHub Actions on its own bare-metal hardware. Instead of the general-purpose servers behind most cloud runners, it uses high single-thread gaming CPUs - the kind prized for raw speed - and pairs them with a software stack written specifically for CI. Every job boots into an isolated Firecracker microVM in under three seconds. The pitch is deliberately unglamorous: change one line in your workflow file, keep everything else, and watch your builds finish roughly twice as fast at close to half the cost.

The founders did not arrive at this from theory. Aditya Jayaprakash, Aayush Shah and Aditya Maru met at the University of Waterloo and went on to build large-scale distributed systems at Faire and Cockroach Labs. At Faire, Jayaprakash - who goes by JP - watched how directly build and deploy speed tied to the business. "Every minute of downtime is a minute of lost revenue," he recalled of that period. Conversations across the industry kept surfacing the same complaint: fast compute was hard to get instantly, pricing was unpredictable, and no one could see where the time actually went.

That framing matters. Blacksmith is not trying to build a prettier dashboard on top of someone else's cloud. It owns the layer that competitors rent, which is what lets it undercut GitHub's runner pricing by 60% or more. As the team puts it, if your bread and butter is compute, it makes sense to control the metal it runs on.

Blacksmith's growth has come mostly through developers telling other developers - the kind of adoption that is hard to buy and hard to fake. Its customers are startups and scale-ups that live in their pipelines: infrastructure companies, AI teams, and product shops where a slow CI queue is felt across the whole engineering org.

Blacksmith enters at the exact point where teams feel pain daily and switching is cheap: the CI runner. That wedge - one line of config, immediate speed - gives it a foothold that broader platforms struggle to defend. From there, its roadmap points toward a fuller CI cloud spanning compute, observability and, with Sandboxes, on-demand isolated environments.

The timing lines up with a broader shift. As AI-assisted development pushes more code through pipelines faster, the cost and speed of CI become harder to ignore. Blacksmith's fundraise pitch leaned on exactly that: fast CI as a way to unblock modern, high-velocity software teams. Whether it grows into the category-defining CI cloud its founders describe is still an open question - but the traction, and the fact that a serious investor wrote a second check before the first was spent, suggest the problem is real.