It’s been about three months since Docker Inc. sold its Enterprise business to Mirantis, and the dust is starting to settle. The company that kicked off the container revolution is now a much smaller operation, focused squarely on Docker Desktop and Docker Hub — the developer-facing tools rather than the orchestration platform. Having watched Docker’s rollercoaster journey from the beginning, I think this pivot, while painful, might be exactly what the company needed.
How We Got Here#
The Docker story is a cautionary tale about the gap between technology adoption and business viability. Docker made containers accessible to millions of developers. Before Docker, containers existed (LXC, Solaris Zones, FreeBSD Jails), but they were the domain of systems administrators with deep Linux kernel knowledge. Docker wrapped the complexity in a developer-friendly CLI and image format, and the industry was never the same.
But Docker the company struggled to monetize Docker the technology. The open-source container runtime became a commodity almost immediately. When Docker tried to build an enterprise platform — Docker Swarm for orchestration, Docker Enterprise Edition for management — Kubernetes had already won that battle. Google’s backing, the CNCF ecosystem, and the sheer momentum of the Kubernetes community made Docker Swarm a hard sell to enterprise buyers.
The November 2019 deal saw Mirantis acquire Docker Enterprise, including Docker Swarm, Docker Trusted Registry, and the enterprise sales team. Docker retained Docker Desktop, Docker Hub, and about 300 employees. Scott Johnston took over as CEO with a mandate to focus on developers.
Docker Desktop as the Core Product#
If you’re a developer working on a Mac or Windows machine, Docker Desktop is probably already part of your daily workflow. It’s one of those tools that’s become so ubiquitous that you almost forget it’s a product someone has to build and maintain. Running Linux containers seamlessly on non-Linux operating systems is genuinely hard engineering — the Linux VM management, filesystem sharing, networking, and the integration with local tools all have to work together smoothly.
Docker Desktop currently handles this remarkably well. The switch from the older Docker Toolbox (which used VirtualBox) to the native hypervisor integration (HyperKit on Mac, WSL 2 on Windows) was a significant improvement. On my MacBook Pro, starting a container takes seconds, and volume mounts — while still not as fast as native Linux — are workable for development.
The question is whether Docker Desktop alone can sustain a company. The current licensing model keeps it free for individual developers and small teams, but Docker has started signaling that enterprise features will come with a price tag. Features like vulnerability scanning, team management for Docker Hub, and enhanced security controls for Docker Desktop could form the basis of a subscription model.
Docker Hub’s Critical Position#
Docker Hub is Docker’s other crown jewel, and it’s in a stronger position than many people realize. With over 6 million repositories and billions of image pulls, Docker Hub is the de facto distribution mechanism for container images. Even if you run Kubernetes in production and never touch Docker’s runtime, there’s a good chance your base images come from Docker Hub.
This network effect is powerful but fragile. GitHub Packages is growing, Google Container Registry and Amazon ECR serve their respective clouds well, and projects like Harbor provide self-hosted alternatives. Docker Hub’s advantage is inertia and the sheer volume of existing content — every major open source project publishes official images there.
The recent introduction of rate limiting on Docker Hub pulls has been controversial but economically necessary. The infrastructure costs of serving billions of image pulls are substantial, and Docker needs revenue to survive. I expect we’ll see more tiering and paid features around Docker Hub in the coming months.
The Container Runtime Question#
Here’s an interesting technical subplot: Kubernetes is moving away from Docker as a container runtime. The Container Runtime Interface (CRI) allows Kubernetes to use containerd or CRI-O directly, bypassing the Docker daemon entirely. This doesn’t mean Docker images stop working — they’re all OCI-compliant — but it does mean that in production Kubernetes clusters, the Docker daemon is increasingly unnecessary overhead.
For Docker Inc., this is actually fine. Their developer tools don’t depend on being the production runtime. As long as the image format remains standard (and it will — OCI has seen to that), Docker Desktop can be the place where developers build and test images, regardless of what runs them in production. It’s a narrower niche than “the entire container platform,” but it’s a defensible one.
What concerns me slightly is the growing gap between development and production environments that this creates. If you’re building images with Docker but running them with containerd, there can be subtle differences in behavior — particularly around networking, storage drivers, and security contexts. In practice these differences are rare, but they exist, and they can make debugging production issues frustrating.
My Take#
I’ve been using Docker since the 0.x days, back when it was still based on LXC and every other build broke something. The technology was transformative — it fundamentally changed how we think about packaging and deploying software. But the business always struggled with the classic open-source dilemma: how do you capture value from a technology that everyone uses but no one wants to pay for?
Docker’s new focus is narrower but more honest. Developer tooling is a legitimate business — JetBrains has proven that developers will pay for tools that make them productive. Docker Desktop is genuinely useful, Docker Hub is genuinely critical infrastructure, and there’s a real opportunity to build a sustainable business around them.
The risk is that Docker tries to boil the ocean again — expanding into CI/CD, deployment, monitoring, or other areas where they’d face entrenched competition. The best thing Docker can do right now is make Docker Desktop the absolute best local development experience for containers, and make Docker Hub the most reliable and secure image registry. Do those two things well, and the business will follow.
For the rest of us, the lesson is simpler: Docker the company may change shape, but containers aren’t going anywhere. OCI standards ensure that the ecosystem survives regardless of any single vendor’s fortunes. And that’s exactly how infrastructure technology should work.
This post is part of my Infrastructure Notes series, tracking the tools and platforms that shape how we build and deploy software.