Chapek at HumanX Amsterdam 2026

Three days on the floor at HumanX, talking to founders, infra leads and researchers about what dedicated GPU servers on EU territory look like when everything is in the rate.

HumanX brought together the people building and buying AI infrastructure in Europe. We went with one message: a single tenant per server, eight H200 or B300 GPUs, storage, egress, AI Studio and 24/7 support included, onboarding from December 2026.

Most of the questions were practical. How fast can we start, who else is on the machine, what happens to our data. The answers are short: hours not months, nobody, it stays on EU territory.

What we heard

One visitor read the banner, asked whether we really had H200 and B300 servers, and when we said yes, answered: "I need 2,000." That was the most direct version of a conversation we had many times over the three days.

Most of the floor was about agents and copilots. The questions at our counter were about something else: how to scale, how to make a cloud bill predictable, what it costs to move petabytes out of a hyperscaler, and whether GPUs are actually available when you need them.

Who stopped by

Founders, infrastructure leads and research teams, most of them from Europe, a few from the US and India. The same three questions came up in almost every conversation: when can we start, what exactly is in the rate, and where does the data sit. The answers are short, and they are the reason the booth looked the way it did.

Thank you to everyone who stopped by. If we spoke and you want to pick the conversation back up, the fastest route is below.