The Fortune Was in the Phone Book

Some startups win by inventing something new. A quieter kind wins by noticing a small missing layer on top of something that already exists — a niche everyone else walked past — and building exactly that layer, well. Tailscale is one of the cleanest examples I know.

The protocol beneath was already there, and free

Underneath Tailscale sits WireGuard, an open-source protocol released in 2016 by Jason Donenfeld. It is a small piece of work — about 4,000 lines of code, where the older VPN systems ran to a hundred thousand or more. It makes one fixed choice of modern encryption and refuses to negotiate, which sounds stubborn and is precisely why it is fast and hard to attack. It connects in a single round trip, survives your phone jumping from Wi-Fi to cellular, and says nothing at all to anyone who does not hold a valid key.

WireGuard was free for anyone to use. It still is. But on its own it asks you to copy keys between machines by hand and keep track of which machine may talk to which. For one person with three computers, that is an afternoon. For a company with three thousand laptops, it is a job nobody wants.

Gamers had the ancient version

The desire is older than the protocol. In 2004, a tool called Hamachi let gamers make their home computers believe they were on the same local network, so games built for a roomful of friends worked across the internet. Millions used it. It proved, fifteen years early, that ordinary people wanted private little networks that simply set themselves up. Hamachi was bought, neglected, and faded. ZeroTier kept a version of the idea alive for technical users. The need never went away; it just waited for a better foundation and a bigger moment.

Ex-Google engineers, and the moment

Tailscale was founded in Toronto in 2019 by Avery Pennarun, David Crawshaw, and David Carney — engineers out of Google, joined a year later by Brad Fitzpatrick. It launched in April 2020, the exact month the world’s offices emptied into spare bedrooms.

To be fair, the old way did not collapse. At a big firm — tens of thousands of employees, several VPN gates — remote work ran fine, and it still does. What COVID changed was the shape of the problem. Work spread across home networks and cloud servers, and the friction landed on the people the gate model never fit: the developer who needs to reach one cloud machine, not the whole office network; the small team with nobody to run a gateway; the engineer whose laptop, phone, and home server should simply find each other. For them the traditional answer was heavy — buy a gate, guard it — and it trusted too much, because once you pass the gate you are inside. Tailscale’s answer is finer: each device proves who it is, and may reach only the machines it is allowed to reach.

The key contribution: a phone book, not a tunnel

Here is the part worth studying. Tailscale did not build a better tunnel — WireGuard was the tunnel, and it was free. What they built was the coordination layer on top of it: the phone book.

Their servers know who you are (you sign in with your Google or work account), which devices are yours, and who is allowed to reach what. They introduce devices to each other and swap the public halves of their keys. Then they step aside: the traffic itself flows directly between your devices, encrypted, never through Tailscale. If a firewall blocks the direct path, a relay passes the sealed packets along, still unable to read them. Names replace numbers, keys rotate on their own, and a new laptop joins the network in the time it takes to sign in.

Because they never carry the traffic, the company stays light: roughly 300 people serve millions of users. What they sell is not the road. It is the phone book, the introductions, and the rules — the admin layer nobody found glamorous.

The business that followed

The model is freemium and bottom-up: free for individuals, per-seat pricing for companies, spread by engineers who loved it rather than by a sales force — famously, they grew with almost no marketing. Paying business customers went from about 5,000 in early 2024 to nearly 40,000 by mid-2026. Investors have put in about $275 million; the April 2025 round valued the company around $1.5 billion. Revenue is private and only estimated — roughly $45 million in 2025, with the CEO saying sales have doubled every year.

The complaints are real and worth naming. Per-seat pricing stings small teams. The phone book is run by one company, which sees device metadata (never traffic content) and whose outages freeze logins and changes, though live connections keep working. And there is an escape hatch: Headscale, an open-source reimplementation of just the phone book, for people who want to keep that book themselves. Tailscale tolerates it — they even hired its creator. The skeptic’s question also stands: a $1.5 billion value on estimated tens of millions in revenue asks the future to arrive on schedule.

The protocol was a public good, free to all, for years. The fortune was not in the tunnel; it was in the thin, unglamorous layer above it — identity, introductions, rules — the piece that turned a powerful tool for experts into a product for everyone. The open thing underneath creates the possibility, and the company that wins is often the one that finds the small missing layer everyone needs and nobody wanted to build. The niche is the business.

Leave a Reply