Blacksmith raised $45 million in a Series B led by Peak XV Partners, valuing the AI code-validation startup at $550 million.
Coding agents have turned pull requests into a factory line that wasn’t pre-sized for volume. “Writing code has gotten dramatically easier. Validating it hasn’t,” said co-founder and CEO Aditya Jayaprakash, and the new capital, joined by existing backers Y Combinator and GV, is aimed squarely at that bottleneck: an order-of-magnitude expansion of the compute Blacksmith runs CI workloads on.
Blacksmith’s own new product, codesmith, diagnoses CI failures and auto-fixes pull requests. That’s the same territory CodeRabbit, a separate AI code-review startup, was defending with the $143 million raise it announced a day earlier at a $1.5 billion valuation.
It looks like Blacksmith picking a fight with the review layer. It isn’t: codesmith works inside Blacksmith’s own validation loop rather than on top of it, and the two companies are converging on the same premise; one CodeRabbit’s incoming board member named outright, that AI-generated code needs an “independent governance layer” no matter which model wrote it.
Blacksmith’s customer base has grown from roughly 800 companies at its Series A last September to more than 6,000 today, including Supabase, Clerk, and Mercury.
Sources: DealStreetAsia (Blacksmith), Reuters (CodeRabbit)
What the Research Implies: a Cost-Structure Inversion
Blacksmith and CodeRabbit aren’t just raising capital; they are pricing a structural deficit. AI writes code fast, but the once promised labor reduction cost is now proving to be m more expensive than expected.
Going through a 2026 Google study (Tran et al.) analyzing 3.52 million C++ changes in production, AI triggers a massive cost-structure inversion. It collapses authoring time but spikes chronic maintenance and runtime costs. The study found AI-generated code carries a 1.15x higher “interface and coupling burden” and a 1.39x higher “copy and allocation overhead” than human-written code, driven by a bias toward explicit local loops over optimized standard libraries.
The downstream result is a 5% to 8% increase in compute and memory consumption. Human reviewers struggle to catch this chronic inefficiency, generating 1.92x more blocking threads but ultimately letting the structural bloat pass into deployment.
You can’t govern that bloat if you can’t even see it, as I covered before in Capability Without Visibility: What You Pay for Generative AI Workspaces Is More Than Money.
A 2026 World of Code census (Khosravani & Mockus) spanning 180 million repositories revealed that AI agents are deeply embedded but epistemically camouflaged. Audits relying solely on bot-account activity undercount AI prevalence by up to 30x. Claude Code, for instance, registered 28,154 bot commits, but 850,157 actual commits when triangulated with message signatures.
Furthermore, these camouflaged agents do not care about your security posture by default. A 2026 Vanta study (Sethi & Errico) demonstrated that unprompted models write functional but non-compliant code, scoring 47% to 88% on SOC 2 conformance and leaving out critical infrastructure hardening.
Yet, adding a single sentence demanding SOC 2 compliance shifted scores to 86% to 100% and removed severe vulnerabilities like reachable debuggers. The model knows how to secure the code. It just won’t.
Generation is solved. Governance is the new bottleneck.
Sources:
- Characterizing the Quality Profile of AI-Generated C++ in Production
- Can AI Write Compliant Code, and to What Extent? Evaluating SOC 2 Compliance of Claude Fable 5, Claude Opus 4.8, and Claude Opus 5 Across Four Use Cases
- Detecting AI Coding Agents in Open Source: A Validated Multi-Method Census of 180 Million Repositories
Where the Rest of India’s Tech Capital is Landing
Blacksmith’s $45 million is not the only interesting number here. Peak XV Partners, the same firm that led it, just closed $1.3 billion across new India and Southeast Asia vehicles.
Accel, the global venture firm behind Swiggy and Freshworks, raised $550 million for its ninth India fund the same week, part of a coordinated $3.5 billion global raise. Zoom out further: Moneycontrol tracked ten India-focused funds worth $3.2 billion announced in just the first half of 2026, with Peak XV, Elevation Capital and B Capital alone accounting for nearly two-thirds of that total. One code-validation startup’s Series B turns out to be a single transaction inside a capital wave three orders of magnitude larger.
Not every dollar chasing that wave is betting on what Blacksmith sells, though.
Peak XV managing director G.V. Ravishankar has argued India’s opening isn’t in outspending frontier labs. It’s in the infrastructure, software and tooling layer built on top of them, deployed at a fraction of the capital those labs require.
That’s the same wager underwriting Blacksmith’s raise, just stated as thesis instead of balance sheet: value concentrates not in the model, but in what it takes to make the model’s output usable.
Sources: DealStreetAsia, TechCrunch, Moneycontrol, Livemint
Aklatan’s news and analysis drills down to the structural mechanics, geopolitical shifts, and hidden constraints truly driving AI and Asian tech ecosystems and knowledge work.
See coverage span here: Aklatan’s News and Analysis
Generative AI Transparency:
This news article was written primarily with generative AI, specifically SupraGraphos’ A.C.E. News Module. Reviewed with human post-editing, all sources and claims are confirmed as of the time of writing.