The short version
Blacksmith, a startup that runs continuous integration (CI) builds and tests on its own hardware, raised $66 million at a $1 billion valuation on October 21, 2026, two months after a $45 million round at $550 million. Its weekly CI jobs grew 18-fold this year as AI coding agents push more code into review. For any team using AI to write code, build and test capacity now limits delivery speed, so treat CI as a core part of your cloud and DevOps setup, not an afterthought.
What did Blacksmith announce?
On October 21, 2026, Blacksmith said it had closed a $66 million Series C led by Sequoia Capital at a $1 billion valuation, with Premji Invest joining existing backers Peak XV Partners, GV and Y Combinator. The round came about two months after a $45 million Series B led by Peak XV at a $550 million valuation. TechCrunch reported that the company was valued at $60 million when it raised its $10 million Series A a year earlier. Fortune, which broke the news, put total funding at roughly $125 million.
Blacksmith was founded in 2024 by Aditya Jayaprakash, Aayush Shah and Aditya Maru, all University of Waterloo graduates. It sells CI: the automated builds and tests that every code change must pass before it is merged and deployed. Rather than renting general-purpose cloud servers, the company buys its own machines and runs them in data centers in the US and Europe. In its announcement, it said its fleet is just under 10 megawatts and that it plans to grow it about five times in 2027. According to Fortune, revenue is now in the tens of millions of dollars, its largest customers spend more than $1 million a year, and its customer count roughly doubled from about 5,000 in August to more than 10,000.
Why is CI becoming the bottleneck for AI coding?
Jayaprakash told Fortune that AI tools have sped up code writing, and “now the bottleneck has shifted to the second piece, which is: how do I actually validate all this code?” He told TechCrunch that most companies are struggling to keep up with the volume of CI they need to run. The math is simple. A coding agent can open a pull request in minutes, and each pull request, plus every follow-up commit, triggers a full build and test run. When agents work in parallel, the number of runs grows much faster than the size of the engineering team.
CI load is also spiky. Blacksmith describes demand that swings from zero to tens of thousands of CPUs. Hosted runners that cost little at human pace become a large budget line at agent pace, and queues form at peak hours. Slow feedback hurts twice. Developers and agents wait for results, and long pipelines encourage people to skip or batch tests, which is the opposite of what you want when more of the code was written by a model and reviewed by fewer people.
What it means for US & EU software teams
First, CI is now part of your AI productivity budget. If you adopt coding agents and measure only how many pull requests they open, you will miss the cost that comes after: build minutes, test infrastructure and reviewer time. Track CI minutes and cost per merged pull request before and after an AI rollout. That shows the real gain and stops a cheap coding tool from turning into an expensive pipeline.
Second, test quality matters more than test volume. When a model writes the code, your test suite is the main proof that it works. Flaky tests, which pass or fail at random, waste more money at high volume and train people to ignore red builds. Fixing or quarantining flaky tests, running only the tests affected by a change, and caching build outputs usually save more time than faster hardware. If you run CI on GitHub Actions, dependency caching, path filters and concurrency limits that cancel outdated runs are the first settings to check.
Third, a faster CI vendor is one option, not the only one. Specialist runner providers like Blacksmith compete on speed and price against hosted runners and self-managed fleets on your own cloud account. Each choice moves your source code and build secrets to a different place, so check where jobs run and how credentials are handled. EU teams should also confirm data location if builds touch production data or customer test fixtures, which they ideally should not.
How should teams fix a slow CI pipeline?
- Measure first. Record median and 90th-percentile pipeline time, queue time, and cost per merged pull request.
- Kill flaky tests. Find tests that fail and then pass on retry, and fix or quarantine them.
- Run only what changed. Use path filters and test selection so a docs edit does not trigger the full suite.
- Cache aggressively. Cache dependencies, Docker layers and build outputs between runs.
- Cancel stale runs. When a new commit lands on a branch, stop the run for the previous one.
- Set limits for agents. Cap how many concurrent pull requests and CI runs an AI agent can start.
- Compare runner options. Price hosted, specialist and self-managed runners on your real workload, including security review.
Frequently asked questions
What is Blacksmith?
Blacksmith is a US startup founded in 2024 that runs continuous integration (CI) jobs, the automated builds and tests code must pass before merge, on its own machines in US and European data centers. On October 21, 2026 it raised a $66 million Series C led by Sequoia at a $1 billion valuation. It says more than 10,000 companies use it, including Granola, Handshake, LangChain and Mercury.
Why does AI-generated code put pressure on CI?
Every pull request and every follow-up commit triggers a build and test run. AI coding agents can open many pull requests in parallel, so the number of CI runs grows much faster than headcount. Blacksmith says its weekly CI jobs rose from 5 million in January 2026 to more than 90 million. More runs mean longer queues, higher compute bills and slower feedback unless the pipeline is tuned.
How can a team cut CI time and cost?
Start by measuring pipeline time, queue time and cost per merged pull request. Then fix or quarantine flaky tests, run only the tests affected by a change, cache dependencies and build outputs, cancel outdated runs when new commits land, and cap how many runs AI agents can start at once. Only after that compare hosted, specialist and self-managed runners on your real workload.
Is it safe to run CI on a third-party runner provider?
It can be, but it is a security decision. CI jobs see your source code and often hold deployment secrets, so check where jobs run, how secrets are injected and isolated, how long logs and caches are kept, and what certifications the provider has. EU teams should confirm data location and avoid using production or personal data in test fixtures.
Sources
Blacksmith — Building the CI cloud: our Series C (primary source, October 21, 2026)
Fortune — AI code-testing startup Blacksmith hits $1 billion valuation as AI-written code surges (October 21, 2026)
TechCrunch — Blacksmith hits unicorn valuation after Sequoia leads Series C (October 21, 2026)