No investors, no forms, no brochure
We lit five sites in 2019, deleted our cheapest plan in 2021, and have never taken a euro from anybody who was not buying a server. That is most of the story. The rest is below.
It started with a spreadsheet and four quotes for rack space
The domain was registered in 2012 by somebody with different plans for it. We bought it in 2019 and took the first order that November.
Everyone here came from somewhere that sold oversold cores. You learn the trade quickly: buy the cheapest silicon that will boot, divide it eight ways, advertise the division as a core, and hope the customers who notice leave quietly. Financially it works. Operating it is miserable.
Five sites went live in the first year: Amsterdam, Frankfurt, London, Bucharest and New York. Amsterdam came first and still carries the most instances of anywhere in the fleet. We paid for the racks out of savings, sold servers, and spent the money on more racks. That has been the entire financial model ever since.
Nothing about the early platform was good. Invoices came out of a script running on a laptop, provisioning involved a human reading an email, and Bitcoin was the only asset we accepted because it was the only integration we had finished. The changelog starts in November 2019 and does not flatter anybody.
Why nothing here is cheap
We used to sell a €4 plan. Withdrawing it was the best decision this company has made, and it cost us roughly a quarter of our revenue for three months.
The cheap tier was nine out of ten support tickets. It was every abuse report we received. Both facts held for two years and neither improved with tighter rules, better filtering or a stricter sign-up flow, because the problem was never the customers. Four euros a month buys a machine that is worth burning.
What replaced it is a floor rather than a ladder. The smallest thing we sell is 4 dedicated Zen 5 cores with 16 GB of DDR5 and 200 GB of Gen4 NVMe, at €29 a month. Below that we sell nothing at all, and there is no plan to build anything.
The second-order effects were larger than the revenue. Support stopped being a queue of people who could not afford to care about their own servers. Engineers got time back. Abuse complaints fell far enough that the desk now answers each one properly instead of triaging by volume.
“The cheapest customer is the most expensive one you have.”
Why we never ask who you are
The no-KYC policy is not a stance about privacy. It is a consequence of how we take money, and the two cannot be pulled apart.
Cards drag identity in with them
Accept a card and you inherit a chargeback process, a risk team on somebody else’s payroll, and an obligation to know whose card it is. None of that is optional once the first payment lands. Crypto settlement through OxaPay removes the whole chain, so there is nothing left to verify.
We cannot hand over what was never written down
Sign-up records an email address and a password hash. No address is logged at registration, no name is requested at any point, and no document has ever been asked for at any order value. A request for records gets a short answer because there is a short amount to give.
A threshold would be a lie
Plenty of hosts advertise no-KYC and then want a driving licence the first time an invoice looks unusual. No such threshold exists here. Should one ever be forced on us, the warrant canary stops being signed before the terms page changes.
It is testable, which is the point
Open an account with a throwaway address, pay in Monero, and run an instance for a year. Nobody will contact you about your identity. Nobody here has a mechanism to.
Who owns this, and where the money comes from
No investors, no board, no acquisition, no debt. There has never been a funding announcement because there has never been anything to announce.
The company is owned by the people who operate it. Outside money has been offered four times and declined four times, most recently in 2024, when the term sheet arrived with a request for “identity assurance at onboarding”. That phrase is the entire reason the answer was no.
Hardware is paid for out of revenue, in advance, one rack at a time. It is a slow way to expand, and it is why Johannesburg spent longer racked and waiting than anybody here enjoyed. The alternative is owing somebody a growth curve, and people who are owed growth curves eventually ask for identity documents.
| Source | Share of funding | Notes |
|---|---|---|
| Customer prepayment | All of it | Annual and biennial cycles are why the fleet grows at all. |
| Outside investment | None | Declined four times. The terms always came back to identity. |
| Debt | None | Hardware gets bought when it is paid for, not before. |
| Advertising and data | None | There is no analytics on this site and nothing to sell. |
The practical consequence: if every customer left tomorrow, nothing would be repossessed and nobody could be forced to sell your records, because there is no lender to repossess and no acquirer to sell to.
How decisions get made
There is no product manager. The roadmap is the support queue sorted by how often something comes up, which is less a philosophy than an admission that we never found a better signal.
- 01
Something breaks, or somebody asks twice
Every ticket is read by an engineer who can change the thing being complained about. When the same question turns up a second time in a week it becomes an issue.
- 02
One paragraph, written down
What breaks, who it affects, what it would cost. No document longer than a page has ever survived here, and several have been killed for length alone.
- 03
Two people have to want it
Anyone can veto. Nothing ships because a single person is enthusiastic about it, which has quietly killed more features than any budget ever did.
- 04
It ships or it dies
Work that stalls for a month gets deleted rather than parked. Keeping the backlog short is the only reason a company this size can operate 34 sites.
- 05
It appears on the changelog
Every platform change is published with a date and a version number. If it is not in the changelog, it did not happen.
One standing rule sits above the rest: anything that adds a field to the sign-up form has to survive an argument that nobody has yet won.
What we are deliberately bad at
This list gets longer rather than shorter. Most of it is on purpose.
Talking on the phone
No number, no callback form, no scheduled call with an account executive. Everything happens in writing, where it can be quoted back at us eighteen months later.
Managing your application
We patch a hardening baseline and tell you when it drifts. Your application at 3 a.m. is not something we will debug, and on the few occasions we tried we were worse at it than the customer was.
Selling to procurement
No vendor questionnaires, no security addendum negotiated across six weeks, no signature on a master agreement drafted by somebody else. The terms are one page and identical for everybody.
Being cheap
There is always a host at half our price, and occasionally it is even a good one. We stopped competing on that number in 2021 and stopped pretending to shortly afterwards.
Marketing
No booth, no webinars, no mailing list beyond incident notices. Most people arrive because somebody in a chat channel told them to, which is a growth strategy with obvious limits.
Saying yes quickly
Feature requests get a slow honest answer more often than a fast optimistic one. Two engineers have to agree, and one of them is usually in a bad mood about it.
Questions about the company
The people who run the network. There is no holding company behind us, no investor with a board seat, and no parent we were folded into. We publish no names, which is a deliberate choice for a business whose customers value not appearing on lists.
Fewer than you would guess from 34 sites, more than you would guess from an 11-minute median first response. Everybody who answers a ticket has root, and everybody with root answers tickets.
We do not publish that, and saying so plainly beats inventing a jurisdiction that sounds reassuring. What matters for your data is the law of the site you deploy in. You choose it, and it is printed on every location page.
No. Same operators since 2019, same network, same policy. If that ever changes it goes on the changelog the same day and on the canary page the same month.
None that we will list, because a certification we cannot show you the paperwork for is worth exactly nothing. What we can show you is the incident history, the measured uptime figure and a one-page contract.
A law that applies to us and leaves no alternative. In that case we would say so out loud, stop signing the canary, and give existing accounts notice rather than quietly editing a page at two in the morning.
Nothing here needs a sales call
Pick a site, pick a size, pay in coin. If you would rather argue with an engineer first, the contact page has the address, and the reply comes from somebody who can change the thing you are complaining about.