The bottleneck has moved

With Cursor, Claude Code, Codex, and the agentic tools that have followed, code is now produced at a pace that would have sounded fictional eighteen months ago. A single engineer with a capable agent operates closer to the throughput of a small team. But generating code is only one step of shipping software, and everything that comes after it is still running on infrastructure designed for a world where humans wrote in measured batches. The result is an unobvious but compounding tax on engineering velocity: developers and AI agents alike now spend a meaningful share of every working day waiting on pipelines.

That’s why we’re proud to lead Avrea’s $4.7M pre-seed round, with participation from angel investors, partnering with serial entrepreneurs Hannu Valtonen (co-founder of our unicorn portfolio company Aiven), and Juha Valvanne (founder of Nosto). We’re excited to support the founding team as they build an AI-native CI/CD platform (the infrastructure that helps developers continuously test, integrate, and deliver software) for the era of human and agentic software development. Avrea emerges from stealth today already enterprise-ready, with production workloads live across a growing base of companies.

The default option is showing its age

Most modern engineering teams default to the CI capabilities bundled with their Git provider. It’s the path of least resistance: integrated, “free,” good enough. In the era of AI-generated code, three structural flaws turn into real problems.

It’s slow by architecture. Generic, shared cloud runners on commodity virtual CPUs, so even a relatively simple software build can take far longer than it should: something that could finish in about 90 seconds on optimized infrastructure may take eight minutes instead. That gap looks like a minor annoyance per push, but across hundreds of branches, dozens of agents, and thousands of commits a week, it can add up to 30 to 60 minutes a day, per developer, spent waiting.

It’s expensive by accident. Per-minute pricing on inflated build times creates an awkward incentive: the slower the pipeline runs, the more the customer pays. As AI tooling pushes code volume up by an order of magnitude, that bill scales the wrong way, too.

It was never designed for agents. The dominant CI platforms assume a developer pushes, walks away, and comes back to a result. Agentic workflows (where AI agents iterate, fail, fix, and rerun at machine speed) break that model entirely. The observability is too shallow, the feedback loops are too slow, and the programmatic access is too limited.

The result: Every line of code flowing through the platform becomes the most stagnant part of the modern developer toolchain. It’s commodity, it’s costly, and it’s increasingly in the way.

Avrea: instant value, then a much larger ambition

Avrea is an AI-native CI/CD platform that runs up to 4x faster and at as much as 80% lower infrastructure and execution costs than the dominant alternatives. The hardware is selected for performance. The runtime is engineered, not rented. AI-driven optimizations close the gap on slow tests and flaky pipelines. Observability and debugging are first-class features; these surface the root causes of stuck builds, flaky tests, and resource issues that traditional systems mask behind unclear failures. And switching is straightforward: Avrea runs natively with existing GitHub workflows, ships with ISO certification at launch, and teams feel the difference on the first build.

What caught our attention: a team that’s done this before, and is already shipping

Our conviction in this team is simple: we’ve partnered with them before, seen them execute up close, and know they can pull this off. The proof is already on the table.

Hannu Valtonen co-founded Aiven and, as CPO, drove its product roadmap. He did it by reading a wave early, the migration of open-source data infrastructure into managed cloud services, and shipping the platform that productized and monetized it. Before Aiven, Hannu spent five years as Lead Architect for F-Secure, shipping security infrastructure at industrial scale.

Juha Valvanne brings his own scale story. He is the former CEO and founder of Nosto, the leading commerce experience platform globally. 

The two have known each other for over twenty years. Two co-founders who have each independently taken a B2B platform from zero to category leader. Now teaming up, that combination is rare at any stage and unheard of at pre-seed.

Around them, they’ve assembled a ten-person team of senior operators and engineers, with the majority previously founders, and core hires with experience at Hoxhunt, Spotify, and the broad Helsinki infra ecosystem. In six months from incorporation, this team built a production-grade platform from the ground up, shipped with enterprise security certifications in place, and onboarded more than twenty companies running real workloads, all before the public launch. That kind of velocity, paired with that level of operating maturity, is what infrastructure businesses are made of.

Earlybird General Partner Paul Klemm shares: “Backing Hannu a second time was an easy decision. At Aiven, he built a category-defining infrastructure company and scaled it to unicorn status. But what makes Avrea especially compelling is the opportunity: AI is driving an explosion in code, and the systems that test and ship software are quickly becoming the bottleneck. With Juha and a team deeply experienced in building for developers, Avrea is uniquely positioned to define the future of software delivery.”

 A large and accelerating market

The opportunity is sized appropriately for the ambition. The global CI/CD tools market is projected to grow from ~$9.4bn in 2025 to ~$33.6bn by 2034 (~15% CAGR), within a broader DevOps category, expanding from ~$13bn to over $80bn over the same horizon. Past category leaders have built durable businesses at scale: GitLab IPO’d at $11bn, GitHub was acquired by Microsoft for $7.5bn, CircleCI raised at $1.7bn, CloudBees and Harness sit at $1bn and $3.7bn respectively. Each was built for a previous generation of how software gets shipped. The next platform has to be built for the next AI-native generation.

Why we invested

Earlybird has known and partnered with Hannu since the Aiven days. The relationship is not a pitch we walked into; it’s a multi-year partnership of trust that made it natural to lead this round. Beyond the team, our conviction rests on three things:

• The market shift is real and not yet priced in: CI/CD is already a multi-billion-dollar market, expected to roughly triple over the next decade, yet the current tooling was built for a pre-AI development model. Code volume has exploded; the platforms that integrate and ship it have not kept pace. The first players to rebuild this layer for the era of human and agentic software development will quickly compound advantages.

• The wedge is concrete and already proven: Avrea doesn’t need a customer to believe a future thesis to switch. It needs them to run one build. Twenty production customers before public launch suggest the gap is real and the product is compelling.

• The platform ambition is credible: Lightning-fast CI/CD runners are the entry point. From there, the team will expand the platform beyond runners with deeper observability and AI-powered features that improve build reliability and test quality. Two repeat infrastructure founders, with a senior bench around them, have built and sold to developers at that scale before.

We’re proud to back Avrea, and to do it with founders we’ve watched, partnered with, and trusted for years.

Read more coverage in Tech Funding News and Tech EU.