Infrastructure for the AI Coding Era
In a Nutshell
Blacksmith built a CI-optimized cloud using gaming CPUs and aggressive caching to eliminate the validation bottleneck created by agentic coding tools that let developers write far more code than they can currently ship. Their infrastructure runs on rented gaming servers that evolved into owning megawatts of data center power, delivering 10-15 second job starts versus 15-20 minutes on GCP, which won five-to-six-figure contracts without a sales team. The company scaled to a $1B valuation and 30-person team by keeping the product surface small, building custom fleet management, and preparing for autonomous QA that will validate code changes through artifacts instead of manual review.
These notes were generated by AI and may contain inaccuracies.
Blacksmith builds infrastructure to run CI checks extremely fast. Every time a developer writes code, they must run checks to ensure nothing is broken. The company has grown 20x in the last year and recently raised their Series C at a billion dollar valuation, led by Sequoia, raising approximately 66 million. The founders were outsiders with no background in platform engineering or DevOps, and none had written a CI pipeline before.
The team's approach succeeded partly because they were naive enough not to kill a good idea before it met the market. With agentic coding tools, developers write more code than ever before. However, writing 100 times more code is not useful if developers cannot ship at that speed. The bottleneck has shifted from writing software to validating it, which means waiting on CI to ensure tests pass and build confidence in shipped code.
The founder always wanted to start a company in Silicon Valley. During high school, reading Wikipedia pages about technology companies and their founding stories was a favorite activity. They attended Waterloo for its strong startup program and technical focus. After working at Faire, they began developing ideas more seriously. Blacksmith was their second YC application. The three co-founders were friends from undergrad who were all in Toronto at the time and wanted to start a company together.
All co-founders were engineers who experienced universally poor CI at every company they worked at. CI was slow and managing this infrastructure was consistently painful. They believed they had a better approach. When they first discussed the idea that became Blacksmith, they were initially skeptical about whether it could work. Their conviction grew as they conducted more research. Two co-founders were database engineers, and one worked on search infrastructure.
The founders observed that managing CI infrastructure was difficult because CI workload characteristics differ significantly from what hyperscalers like AWS, GCP, and Azure optimize for. This mismatch caused the pain points. The solution was building a cloud specifically optimized for CI workloads. CI workloads share characteristics with desktop applications, particularly video games running on gaming PCs, which prioritize single-thread execution speed over multiple threads.
Initially, the team rented servers from a German data center services company known for hosting Minecraft and Call of Duty game servers. As scale grew, they began owning more of the stack and now operate at larger agreements with data center providers, purchasing megawatts of power rather than individual machines. An investor noted the unusual approach of using gaming CPUs for CI, as this solution had never been seen before. The concept of purchasing hundreds of gaming CPUs for compilation work was novel compared to typical single-machine gaming setups.
Making CI fast requires skipping redundant work through caching. A significant part of Blacksmith's approach involves moving cache artifacts close to compute nodes.
Acquiring the first 200 customers was extremely difficult. The team initially asked friends at companies to try the product and obtained logo rights to build credibility on their website. They personally called outbound to CTOs, VP of Engineering, and platform engineers. Early customers were small companies that prioritized speed and had minimal bureaucracy. The only way to win over engineers is through performance and solving problems they cannot solve themselves. Engineers are discerning and like trying new tools, though they may not be the stickiest customers.
Without any sales team, Blacksmith secured large five to six figure contracts due to how easily the platform optimized and accelerated CI. The product allows switching all CI to Blacksmith with one click. Langchain, a recent customer, previously ran CI on GCP where developers waited 15 to 20 minutes to acquire compute. On Blacksmith, wait times dropped to 10 to 15 seconds due to effectively infinite compute pools, significantly improving developer velocity.
Blacksmith is executing a massive marketing campaign across San Francisco and New York to position the brand as the immediate choice for engineering teams thinking about CI velocity and code confidence. The campaign includes taking over 15 subway stations in New York and wrapping a subway train in the company's bright yellow color.
When an early team member joined, the company had approximately five people running a small cluster of a couple hundred servers. The team has grown to 30 people with 12 engineers. Key decisions enabling this scaling include developing bespoke software to manage fleet growth, implementing custom observability at every stack layer, and ruthlessly keeping product surface area small and manageable, scaling only with team size rather than accommodating bespoke customer requirements. This focus enables scaling to large numbers of machines and customers with a small team.
A major challenge is keeping up with demand. Every time a new LLM releases, CI usage spikes as teams rush to test and ship more code. Since ordering and installing machines takes time, the team uses technical solutions including better job placement algorithms, more efficient job packing, and improved network topologies. After the Series B announcement, a power surge in Phoenix disabled both primary and backup cooling systems in one of their largest data centers, causing longer job startup times during the outage. This incident highlighted the importance of geographic distribution.
Blacksmith is developing autonomous QA capabilities that gather artifacts like screenshots and video snippets to provide confidence in code changes. With increasing code production volume, developers will lack time to manually verify functionality and will instead seek evidence. Blacksmith aims to provide this confidence to enable faster code merges, removing the validation bottleneck and allowing developers to ship software at the speed of their thoughts.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
