Docker, Inc.

01Who They Are Now

Open a laptop in any engineering office in 2026. Within ninety seconds, a whale logo lights up the menu bar.

It is a Tuesday morning at a fintech in Berlin. A developer types docker compose up and walks to the espresso machine. By the time she returns, a Postgres database, two microservices, a Redis cache, and a small open-weight language model are running on her laptop. None of it touches her host system. None of it cares whether her colleagues in Singapore use Windows or her CTO uses a Framework laptop running NixOS.

This is Docker, Inc.'s product. Not the whale - the boredom. The unspectacular fact that software now packs, ships, and runs the same way everywhere, and almost nobody talks about it anymore. That is, by every reasonable measure, an extraordinary outcome for a company that started life as a struggling French platform-as-a-service startup named dotCloud.

Roughly twenty million developers use Docker tools every month. Docker Hub, the public registry the company maintains, serves billions of image pulls. The Fortune 100 pays for Docker Business. And the new CEO, Don Johnson - hired in February 2025 from Oracle Cloud Infrastructure - says the next decade looks a lot like the last one, except with model weights inside the containers.

02The Problem They Saw

For the better part of three decades, shipping software was an act of polite lying. You promised your code would run on your colleague's machine. It rarely did. You promised it would run in staging. Sometimes it did. You promised production would behave like staging. Production laughed.

The industry's response to this comedy was to build heavier and heavier scaffolding around it. Virtual machines. Configuration managers. Hand-rolled bash scripts maintained by exactly one person who is now on parental leave. The result was an entire profession - DevOps - whose primary job was apologizing for the gap between "it works on my machine" and "it works."

Solomon Hykes, the French engineer who co-founded dotCloud in 2008 with Sebastien Pahl and Kamel Founadi, watched this gap from inside it. dotCloud was a PaaS - a platform for running other people's code - and the team had built a small internal tool to make deployments more predictable. It wrapped Linux containers, a kernel feature that had existed for years, in a usable command-line interface. They almost did not release it.

03The Founders' Bet

In March 2013, Hykes walked onto the stage at PyCon in Santa Clara and gave a five-minute lightning talk titled "The future of Linux containers." The slides were unfussy. The demo was short. The reaction was not.

Within seven months Docker had been downloaded one hundred thousand times. Within a year it had become one of the fastest-growing open-source projects in the history of GitHub. By 2015, after a $95 million Series D led by Insight Venture Partners, the company was valued north of a billion dollars. It had achieved the status now politely called unicorn and, more importantly, the status engineers actually care about: it was the default.

The bet underneath all of this was a counterintuitive one. Give away the engine. Donate the runtime - runc, then containerd - to neutral foundations. Refuse, repeatedly, to be the company that controls the standard. Build the business on the workshop above the engine: the developer tools, the registry, the cloud builds, the security scans, the things humans pay for because their time is more expensive than their software.

It did not always work. In 2019, after years of trying to sell Docker Enterprise into the same enterprises that had grown comfortable using Docker for free, the company sold that enterprise unit to Mirantis and reinvented itself. The press wrote obituaries. The developers kept typing docker run.

04The Product

Strip Docker's catalog down and what remains is a single, stubborn idea: a container is a contract. A contract about what your code expects, what it depends on, and how it should behave. Write the contract once - in a Dockerfile, a Compose file, a Hub image - and you can hand it to anyone, anywhere, including a robot.

That contract is sold today in a small constellation of products. Docker Desktop is the front door - the GUI that runs containers on a Mac, a Windows box, or Linux. Docker Hub is the warehouse - the largest public registry of container images on the planet. Docker Build Cloud rents out a shared, remote build cache so that engineers waiting for Docker images to compile have less time to read social media. Docker Scout watches your supply chain. Docker Hardened Images give security teams a starting point they do not have to apologize for.

And then, newer: Docker Model Runner, which packages open-weight language models so they boot the same way a Postgres container does, and the Docker MCP Toolkit and Catalog, which give AI agents a registry of context-protocol servers they can actually find. The bet is that every AI workload, eventually, is just another container with a slightly more dramatic README.

05The Proof

It is a strange thing to evaluate a developer-tools company. The customers are mostly invisible. They show up in conference talks, GitHub stars, and Slack channels - not glossy logo walls. So you look at the indirect evidence.

The indirect evidence: Docker Hub serves billions of image pulls a month, on a network that has to behave whether a startup in Lagos is fetching a base image or Capital One is pulling a hardened build. Microsoft and Docker have shipped joint Windows integrations since 2014. AWS has noted that more than 80 percent of its cloud customers run containers in some form. GitHub Actions, Atlassian Cloud, and NVIDIA all wire into the toolchain.

On the financial side, the company is private and quiet, but estimates put annual revenue near $207 million, with growth driven almost entirely by subscriptions paid by organizations that started using Docker for free and eventually realized their engineers refused to live without it. That is a particular kind of moat. It is also one of the only kinds that compounds without sales reps.

06The Mission

The official line is short: help developers build, share, and run any application, anywhere. Posted on a wall in San Francisco. Reposted, somewhere, by an intern in a thread about company values.

The unofficial line is shorter. Make shipping software boring. Make the gap between an idea and a running system small enough that a single engineer can close it in an afternoon. Treat the work of packaging - which sounds clerical and dull and is, in fact, the entire game - with the seriousness it deserves.

Mark Cavage, the new president, joined in March 2025 from Oracle Cloud, where he had spent years on the unglamorous infrastructure that powered other people's products. He has been blunt about where the puck is going: containers, in his view, are the most natural packaging unit for AI workloads we have. Not because they are perfect, but because they are familiar - and familiarity, at scale, is worth more than novelty.

07Why It Matters Tomorrow

Return to that developer in Berlin. The one with the espresso and the seven services running on her laptop. In 2014, that demo would have taken three afternoons. In 2019, it would have taken an hour. In 2026, it takes ninety seconds. By 2030, it will include a fine-tuned model, a vector store, three agents arguing about the right approach to a pull request, and a billing integration she did not write. All of it - if Docker is right - inside containers she can hand off, version, and throw away.

That is the bet. Not that AI will be enormous - everyone is making that bet, badly - but that AI will be normal. That it will be deployed, debugged, and forgotten about by the same engineers who deploy, debug, and forget about web apps today. And that they will reach for the same whale-shaped menu-bar icon to do it.

The Docker, Inc. of 2026 looks, from the outside, like a company quietly winning. New CEO. New COO. New products. Old, durable habits in millions of developer terminals. It is not glamorous. It was never going to be. The point of plumbing is that you stop noticing it.

Which, if you have ever shipped software that worked the first time on someone else's machine, is the highest praise a piece of infrastructure can earn.