Blacksmith has raised $66 million in a Sequoia-led Series C at a $1 billion valuation, roughly two months after its previous funding round. The company’s October 21, 2026 announcement pairs that financing with a striking usage claim: weekly continuous integration jobs have risen from 5 million in January to more than 90 million. TechCrunch reports the round size, while Blacksmith’s announcement supplies the weekly figures.
The valuation is the headline. The more consequential question is what happens when software teams can produce changes faster than their infrastructure can validate them.
Blacksmith argues that AI coding agents are accelerating that shift. Its growth supports a narrower, concrete observation: companies are sending substantially more work through its CI platform. A separate engineering account from Anthropic shows how rising code output can strain both the machines executing tests and the systems deciding which tests should run.
Funding and Weekly CI Volume Climb Together
Blacksmith co-founder and CEO Aditya Jayaprakash announced the Series C on October 21. Alongside Sequoia, participants include Premji Invest and existing investors Peak XV Partners, GV and Y Combinator. Blacksmith says the financing brings its total funding to $125 million.
The company runs infrastructure for CI, the automated building and testing of code changes before they are merged. Its customers include Mercury, Granola, Handshake and LangChain, according to the announcement.
| Metric | Published figure | Important context |
|---|---|---|
| Series C | $66 million | Round size reported by TechCrunch |
| Latest valuation | $1 billion | Up from $550 million about two months earlier |
| Previous round | $45 million Series B | Reported by TechCrunch |
| Total funding | $125 million | Confirmed in Blacksmith’s announcement |
| Weekly CI jobs | More than 90 million | Up from 5 million in January, according to Blacksmith |
| Annual recurring revenue | 17-fold growth | Since January; no absolute ARR figure disclosed |
The weekly denominator matters. TechCrunch describes Blacksmith as having run more than 90 million jobs, but the primary announcement is more specific: more than 90 million jobs each week. That represents more than an 18-fold increase from January, not a cumulative lifetime total.
A CI job also is not interchangeable with a code change or an individual test. One change can trigger multiple jobs, and a job can execute many tests. The figure measures activity on Blacksmith’s infrastructure, not the amount of software shipped.
These are company-reported revenue and usage figures, without independent verification in the supplied coverage. Neither source discloses absolute ARR or how much of Blacksmith’s workload comes from AI-written code.
I’d read the combination of workload and revenue growth as a stronger business signal than the valuation alone. Paying demand matters. But without the revenue base, pricing changes or customer mix, it is still impossible to calculate revenue per job or determine whether usage and economics are improving together.
Agent-Written Code Pushes Work Into Validation
A coding agent can generate a change quickly, but that change still needs a path into production. A typical CI process has several stages:
- Accept the change: A new or updated pull request triggers checks.
- Prepare the software: Infrastructure retrieves dependencies and builds the relevant components.
- Run validation: Tests and other automated checks look for failures.
- Return results: Developers or agents use those results to revise the change or proceed toward merging.
More changes mean more opportunities to trigger this process. More tests can increase the work associated with each change. Agent retries can add another feedback cycle: generate a patch, run checks, inspect failures and try again.
Anthropic’s engineering account provides a concrete example. The company reports that its engineers ship eight times as much code per quarter as they did during 2021–2025, with Claude authoring 80% of that code. It also reports a tenfold increase in tests and a 25-fold increase in CI jobs over six months.
Those multiples should not be multiplied together to estimate CI demand. Anthropic explicitly notes that every test does not run on every pull request. Test selection, workflow design and the number of iterations all affect the resulting job count.
Anthropic is one documented engineering case, not a representative industry sample or a performance study of Blacksmith. It nevertheless illustrates a plausible mechanism behind Blacksmith’s thesis: faster code production moves pressure downstream, where validation still consumes compute and depends on coordinated systems.
Blacksmith Targets Runner Capacity and Build Overhead
Blacksmith positions its service as a replacement for the runners that execute GitHub Actions jobs. Its product page describes bare-metal gaming CPUs, microVM provisioning, colocated caches and persistent Docker layers.
These mechanisms address different parts of a CI run:
| Mechanism | Intended benefit | What determines its usefulness |
|---|---|---|
| MicroVM provisioning | Reduce the wait for an execution environment | How much time jobs currently spend waiting for runners |
| CPUs with high single-core performance | Shorten CPU-bound work, especially tasks with limited parallelism | Whether compute is the actual bottleneck |
| Caches near the compute | Reduce time spent retrieving cached artifacts | Cache availability, size and reuse |
| Persistent Docker layers on NVMe storage | Reuse unchanged build layers across runs | How often builds can reuse those layers |
The distinction between waiting and executing is important. A faster processor cannot help a job that has not received a runner. More runners cannot eliminate time spent downloading dependencies. Persistent build layers help only when enough of the build remains reusable.
Blacksmith advertises substantial speed and cost improvements, but those comparisons are vendor claims. The supplied evidence does not provide independent measurements or enough benchmark conditions to predict savings for a particular repository.
My view is that teams evaluating this category should first identify where their CI time goes. A workload dominated by queued jobs presents a different buying decision from one dominated by dependency downloads or slow tests. A single headline speedup hides those differences.
Frequently Asked Questions
3 questions
1How much has Blacksmith raised?
Blacksmith has raised $125 million in total, including the latest $66 million Series C led by Sequoia at a $1 billion valuation.
