Docker didn’t set out to reinvent how the internet ships software. It began as a side project inside a struggling Platform-as-a-Service company trying to survive a market that never grew the way its investors had hoped. The container tooling that company built to manage its own servers ended up outliving the platform it was built for by more than a decade, and the pivot behind that reversal is one of the sharper turnarounds in recent tech history. Here’s the version worth knowing: how a lightning talk at a Python conference turned into the packaging format running underneath most of self-hosting, cloud infrastructure, and CI/CD pipelines today.
A Paris PaaS looking for an edge (2008-2012)
dotCloud was founded in Paris in 2008 by Solomon Hykes, Kamel Founadi, and Sebastien Pahl as a Platform-as-a-Service startup, the kind of company that promised developers they could push code and let someone else handle scaling, deployment, and infrastructure. It went through Y Combinator’s Summer 2010 batch, picked up early angel funding from investors including Chris Sacca and Ron Conway, then a $10 million Series A led by Benchmark Capital and Trinity Ventures, and relocated to Silicon Valley. Behind the scenes, dotCloud’s engineers had built an internal tool to package and run application containers consistently across the company’s own servers, wrapping the Linux kernel’s existing cgroups and namespaces primitives into something a developer could use without becoming a kernel specialist first.
A lightning talk that upstaged the platform (2013)
In March 2013, Hykes gave a short lightning talk at PyCon US in Santa Clara, California, demoing that internal tool, now named Docker, and released it as open source under the Apache 2.0 license. What was meant to be a minor technical showcase became the talk of the conference. Docker packaged an application together with its dependencies into a single portable image that ran the same way on a laptop, a test server, or a production machine, using LXC as its original execution driver. Developers who had spent years fighting “it works on my machine” problems got the pitch immediately, and Docker’s GitHub stars started piling up faster than dotCloud’s actual product ever had.
The pivot: betting the company on a side project (2013-2014)
By mid-2013, dotCloud’s PaaS business was struggling to find a market while its internal side project drew more attention than the company itself. New CEO Ben Golub, who joined that July, backed a full pivot: dotCloud, Inc. officially renamed itself Docker, Inc. on October 29, 2013, and about a year later sold the original dotCloud platform and brand to the Berlin-based PaaS provider cloudControl to focus entirely on containers. Docker replaced its dependency on LXC with its own component, libcontainer, and shipped its first production release, Docker 1.0, in June 2014, the point the company considered containers ready for real production workloads instead of demos. That same July, Docker acquired a two-person startup called Orchard, whose small tool for running multi-container apps, Fig, got folded in and relaunched as Docker Compose, the YAML-based workflow homelabbers still reach for by default.
Microservices, funding, and the format wars (2014-2015)
Docker’s timing lined up almost perfectly with the industry’s shift toward microservices, and 2014 brought a wave of partnership announcements from Microsoft, IBM, Amazon, and Red Hat, all racing to support containers on their platforms. The company raised a $95 million Series D round in April 2015 that pushed its valuation past $1 billion. Success also drew rivals. CoreOS, unhappy with the direction Docker’s tooling was heading, launched a competing runtime called rkt, built around its own App Container image spec, threatening to split the young ecosystem into incompatible formats before it had even matured. To head that off, Docker and a broad coalition of vendors, including CoreOS, Google, Microsoft, Amazon, Red Hat, and IBM, founded the Open Container Initiative under the Linux Foundation in June 2015, donating Docker’s own container format and runtime as the starting point for a vendor-neutral standard.
Not your typical venture fund: one of Docker’s early investors was In-Q-Tel, the CIA’s investment arm. It had quietly put money into Docker, a detail that only became public in 2016 and says something about how closely intelligence and defense circles were watching container technology at the time.
Growing pains and the Mirantis deal (2016-2019)
Standardizing the format didn’t solve Docker’s actual problem: the company had built a wildly popular open-source tool without a clear way to turn that popularity into revenue. Kubernetes, originally released by Google and increasingly preferred over Docker’s own Swarm for orchestrating large container fleets, became the de facto standard for running containers at scale, while Docker’s enterprise sales struggled to keep pace with the money it had raised. In 2017, Docker spun its core engine components out into the open source Moby project, separating the low-level container runtime work from the branded Docker product, a reasonable technical move that did nothing for the revenue problem. In November 2019, Docker sold its Enterprise platform business, including the Docker Trusted Registry, Universal Control Plane, roughly 300 of its 400 employees, and 750 enterprise customers, to Mirantis. What remained of Docker, Inc. refocused entirely on the developer-facing side of the business, Docker Desktop and Docker Hub, backed by a fresh $35 million funding round. It read like a retreat at the time. In hindsight, it looks more like the move that saved the company.
A subscription model and a second act (2020-2025)
The refocused Docker, Inc. spent the following years turning Docker Desktop into a real commercial product. In August 2021, the company ended free Docker Desktop access for larger businesses and introduced paid subscription tiers, a change that annoyed plenty of self-hosters even though Docker Engine on Linux stayed untouched. A $105 million Series C round in 2022, led by Bain Capital, funded a run of acquisitions aimed at developer experience and testing: Nestybox, Tilt, Atomist, Mutagen, and AtomicJar. That same year, Kubernetes formally dropped its built-in dockershim support in version 1.24, a symbolic moment more than a practical one, since images built with Docker still run everywhere, but Docker’s own daemon is no longer required to run them. Don Johnson took over as CEO in February 2025, and the company has since leaned hard into AI tooling, building a catalog of Model Context Protocol servers and acquiring MCP Defender to help secure the agent workloads increasingly running inside containers.
Where Docker stands today
Docker never got its PaaS business back, and it never became the sprawling, all-purpose cloud platform dotCloud originally set out to build. What it built instead was the packaging format that container runtimes such as containerd and Red Hat’s daemonless, rootless Podman still speak, the same OCI image spec that Kubernetes clusters pull from no matter which company built the tooling around them. For homelab and self-hosting use, that legacy shows up constantly: management interfaces like Portainer give a web UI to a Docker host, lighter tools such as Dockge exist specifically to manage Compose stacks the way Docker originally imagined developers packaging multi-container apps, and update automation like Watchtower, whose repository was archived in December 2025 after years of quietly keeping containers current, grew out of the same ecosystem of small utilities built to make daily Docker use less manual. If you’re only getting started with any of this, our homelab setup guide covers where Docker fits alongside the rest of a first server build. The platform that almost didn’t survive its own PaaS company’s collapse ended up outlasting most of what came out of that era, one lightning talk at a time.