Chainguard Put Open Source on a Factory Line - and Made Patching Someone Else's Problem

The least glamorous line item in modern software may be the one quietly eating the week. An engineer pulls a public container image. A scanner finds vulnerabilities in packages the application may never use. Security opens tickets. Platform engineering rebuilds the image. Compliance asks for proof. Then a new advisory arrives and everyone takes another lap. Chainguard looked at this ritual and made a blunt commercial proposition: stop treating inherited open-source maintenance as a hobby.

The Kirkland, Washington company supplies rebuilt versions of the open-source components beneath enterprise software. Its containers, language libraries, virtual machines, operating-system packages, CI/CD actions and agent skills are assembled from source inside a controlled factory. They arrive stripped down, cryptographically signed and accompanied by software bills of materials and provenance records. When fixes appear upstream, the factory rebuilds again. The customer buys the artifact and the maintenance promise around it.

The product is the absence of work

Chainguard is easy to mistake for a vulnerability scanner because it lives in the same budget neighborhood. The distinction matters. Scanners inventory and alert. Chainguard changes what enters the inventory. A typical general-purpose image may include shells, package managers and utilities inherited from a broad Linux distribution. Those extras are handy for debugging and irrelevant to the running application - until every disclosed flaw in them becomes somebody's ticket.

Minimal images shrink that denominator. Chainguard's build system then tracks the source, packages the necessary components, generates an SBOM, signs the result with Sigstore and publishes attestations about how it was made. Production subscriptions add remediation commitments: its public plan lists seven days for critical CVEs and 14 days for other severities when qualifying fixes are available. The pitch is operational, not mystical. Fewer unnecessary packages produce fewer findings; controlled builds produce better receipts; continuous rebuilding moves recurring toil to the vendor.

This approach does not secure an application's own code, stop a bad authorization rule or replace runtime defense. It attacks the foundation layer: the huge volume of code an organization did not write but still ships. That boundary is also where Chainguard sits in the market. Docker Hardened Images, Minimus, RapidFort, Echo, Bitnami Secure Images, WizOS, Iron Bank and hardened enterprise Linux offerings are alternatives with different catalogs, build models and compliance packages. A determined platform team can also use open tools such as Wolfi, apko and melange to build its own supply line. The choice is buy, build or tolerate the patch queue.

A four-image test settled the argument

Hapag-Lloyd offers the cleanest account of what this looks like away from a pitch deck. The shipping company was modernizing its Freight Information System, a platform with roots in the early 1990s, from a mainframe monolith into cloud services. Its platform team already automated delivery and treated infrastructure as code. The sticking point was not whether engineers could harden container images. It was whether maintaining a secure-image factory was a sensible use of them.

“We would have needed an entirely new skill set and a dedicated team, with no guarantee we could successfully build something like the factory Chainguard already has.”Johannes Witte, Hapag-Lloyd

The first thing to fail was the build-it-yourself economics. Hapag-Lloyd ran a four-image proof of value, then replaced 25 open-source images with Chainguard versions in one two-week sprint. It paired the vendor with an internal “Update Champion” practice - two people in each sprint responsible for updates and dependencies. Later, the team was using roughly 60 Chainguard images and rolling out hundreds of security updates during busy periods without adding headcount.

What changed minds was visible scanner output. The team reported that hundreds of vulnerabilities across severity levels vanished to zero. Standardized versioning, permissions and update behavior made the collection easier to operate. The question changed from “How do we fix this vulnerability?” to “Are we on the latest version?” That is Chainguard's value proposition in its most useful form: not perfect security, but a dramatically cheaper recurring question.

Another customer example puts a dollar figure on the surrounding compliance burden. Collaboration.Ai used Chainguard with Second Front and AWS while pursuing a federal authorization path for its CrowdVector platform. The companies reported compressing a process expected to take two years into 90 days, reducing vulnerabilities by 97 percent and lowering compliance costs by $2 million. Those outcomes cannot be credited to Chainguard alone - three vendors and the customer's own team were involved - but they show where hardened artifacts become financially interesting: environments where every vulnerability and every missing receipt slows a regulated sale.

The price of making “free” software boring

Chainguard's raw material is open source, but its product is not free. Five container images can be used at no charge. Production images are sold individually by type, while access to the full catalog of more than 2,000 images is licensed by engineering-organization size. Public pricing starts at $19,000 for a team of ten. Libraries are priced by language ecosystem and developer count. Virtual machines use per-image or catalog licensing. FIPS and STIG variants, custom assembly, supported historical versions and enterprise service sit higher in the stack.

That can be expensive for a small team and inexpensive for a regulated enterprise. The honest comparison is not $19,000 versus zero. It is $19,000 versus the loaded cost of engineers rebuilding bases, triaging findings, documenting exceptions, chasing upstream fixes and preparing audit evidence. It also includes migration, testing and the risk of adopting a fast-moving minimal distribution. Chainguard wins when recurring maintenance costs more than the subscription and the artifact catalog covers the workload. It loses when that equation reverses.

Containers were the wedge, AI is the expansion story

The founding story began before the product catalog did. Dan Lorenc and his eventual co-founders - Matt Moore, Ville Aikas, Kim Lewandowski and Scott Nichols - had worked at Google on cloud-native and supply chain projects including Minikube, Distroless, Knative, Tekton, Kaniko, Sigstore and the SLSA framework. Lorenc had already become uneasy about how easily a trusted maintainer could slip malicious code into a popular open-source project. The 2020 SolarWinds compromise made the abstract risk painfully legible.

In 2021, Moore texted Lorenc: “Let's start a new company.” Lorenc gave notice at Google the next day. Chainguard raised a $5 million seed round that December, launched secure base images in 2022 and found its commercial wedge. Financing accelerated with a $50 million Series A, $61 million Series B and $140 million Series C. A $356 million Series D in April 2025 valued the company at $3.5 billion. Six months later, General Catalyst provided $280 million in growth financing, taking total capital secured to $892 million.

The money is funding a broader claim on the software shelf. Containers led to malware-resistant Python, Java and JavaScript libraries, then minimal virtual machines, OS packages, CI/CD actions and agent skills. Chainguard Repository, introduced in beta in 2026, aims to put these artifacts behind one policy-controlled delivery experience. Commercial Builds brings software vendors into the factory. Guardener uses an agent to rewrite legacy Dockerfiles, test the result and show changes in image size, filesystem contents and vulnerabilities.

AI makes that expansion timely and slightly awkward. Coding agents can generate applications and pull dependencies faster than human reviewers can inspect them. Chainguard's partnership with Cursor is designed to steer agentic development toward vetted components. Partnerships with Endor Labs and Upwind connect guarded inputs to dependency intelligence and runtime verification. Yet AI also increases the obligation to prove the factory itself is trustworthy. A machine-generated patch is not useful because it looks plausible. Chainguard says candidate fixes are isolated, tested against the project's baseline and checked with an exploit written before the patch. The unglamorous work remains testing.

What to copy - and when to walk away

The most portable lesson is not “buy Chainguard.” It is to treat inherited vulnerabilities as an operating expense instead of a background annoyance. Pick four heavily used images. Record package count, vulnerability count, remediation hours, audit evidence and breakages. Test cleaner replacements without changing every pipeline at once. If the delta survives a production sprint, expand by service family and assign explicit owners for updates. Hapag-Lloyd's Update Champions matter because even a continuously rebuilt artifact must still be adopted.

The model will not work everywhere. Minimal images may omit shells or utilities that teams quietly depend on. Unsupported packages, legacy binaries and full-distribution assumptions can turn a neat Dockerfile change into a debugging expedition. Weak test suites make frequent rebuilds risky. A company with a narrow stack and strong internal release engineering may build the same controls more cheaply. A startup without compliance pressure may reasonably accept more scanner noise than a bank, defense contractor or global logistics platform.

Chainguard's bet is that these exceptions shrink as its catalog grows and software consumption moves through agents. Its strongest argument is not the octopus mascot, the unicorn valuation or even the zero-CVE label, which is always a snapshot. It is the claim that enterprises should stop downloading anonymous complexity and hiring expensive people to apologize for it later. Put open source on a controlled line, attach the receipt, keep the line moving. For enough companies, boring software may be worth a surprisingly large bill.