About HuneX

A Danish home for your AI infrastructure. We run the hardware. You run the rest.

HuneX owns and operates its own data center on Danish soil - power, cooling, networking and GPU capacity. You keep full control of everything on top: your models, your orchestration, your data.

See pricing Book a walkthrough
OWNERSHIP Danish-owned, Danish-operated. No foreign parent company.
LOCATION Our own data center in Denmark. Not a reseller, not a marketplace.
BUILT FOR Inference and fine-tuning - not the training arms race.
EGRESS 0,00 DKK. Data transfer is never a separate line item.

WHY WE EXIST

Europe rents its intelligence from somewhere else.

The GPU market grew faster than Europe's ability to host it. That gap is the whole reason HuneX exists.

The US CLOUD Act lets American authorities compel US-owned providers to hand over data - even when that data physically sits in an EU data center. Ownership, not server location, decides who can be forced to comply.

Between late 2024 and early 2026, the price of an H100 hour fell by roughly two thirds, and more than 300 new GPU cloud companies entered the market in a single year. Capacity stopped being scarce. Trust did not.

Almost all of that capacity sits outside European jurisdiction. Under the US CLOUD Act, American authorities can compel a US-owned provider to hand over data regardless of which country the servers physically stand in - including servers inside the EU. For a bank, a hospital, a defence supplier or a public authority, that is not a technicality. It is a board-level question with no good answer.

Denmark already has one serious AI compute asset in Gefion, the national supercomputer. But Gefion is built for research and large selected projects, not for a consultancy that needs eight GPUs on Tuesday and two on Friday. And the companies that do sell that flexibility are, with few exceptions, not Danish and not European.

So we built the thing that was missing: commercial, self-service GPU capacity, on our own hardware, in our own building, under Danish law - for the workloads most companies actually run.

<5% of the world's AI compute currently sits in Europe, while US hyperscalers hold over 70% of the regional cloud market.
64-75% drop in H100 rental prices between late 2024 and early 2026. Raw price is no longer a defensible position.
~70% of GPU cloud consumption is expected to be inference rather than training by the end of 2026 (Morgan Stanley estimate).

Where we sit

We are not the cheapest. We are deliberately in the middle.

The GPU market runs from peer-to-peer marketplaces at the bottom to hyperscalers at the top.
Here is the honest map, and where HuneX chose to stand on it.

Cheapest per hour Most expensive per hour
Marketplaces Vast.ai · RunPod ≈ $0.80–2.39 / H100-hour Aggregated third-party capacity. Fastest to start, lowest price, no SLA on community tiers and no fixed jurisdiction. Anyone competing on price alone loses here.
HuneX Danish sovereign mid-market Transparent, all-in, no egress Own hardware, own building, Danish jurisdiction, self-service. Priced against the neoclouds, not against the marketplace floor - because the floor comes with no guarantees attached.
Neoclouds Lambda Labs · Nebius · FluidStack · Sesterce · Oblivus ≈ $1.47–2.99 / H100-hour AI-first infrastructure at serious scale. Strong engineering, but capacity is increasingly committed to a handful of very large anchor tenants - and most of it is not under Danish law.
Hyperscalers AWS · Google Cloud · Azure ≈ $3–7 / H100-hour Broadest ecosystem, highest price, and US ownership regardless of which region you select in the console.
Also on the map

Hetzner and ScanNet

  • Hetzner has owned "cheapest in the EU" since 1997 - but on older GPU classes (A100, RTX), at fixed monthly pricing, with no H100 tier and no spot capacity.
  • ScanNet already runs GPU-as-a-Service from Danish data centers. Denmark is not an empty market, and we do not pretend otherwise.
Our answer

Specialisation, not a broader catalogue

  • We do one thing: AI/GPU infrastructure. No web hosting, no mailboxes, no domains competing for the same engineers.
  • Current-generation GPUs and flexible terms - on-demand or reserved - instead of a fixed monthly server.
  • Pricing published on the website, not "contact us for a quote".

How we work

A clear line down the middle.

Most infrastructure disputes come from an unwritten boundary. Ours is written down, and it does not move.

HuneX owns

The hardware and the uptime

  • GPU hardware, servers, storage and network fabric
  • Power, cooling and physical data center security
  • Capacity planning and hardware lifecycle
  • The uptime commitment in your agreement
  • Danish jurisdiction over everything we host
You keep

Your stack, end to end

  • Models, weights, datasets and prompts
  • Containers, orchestration and scheduling
  • Your own access control and application security
  • Full freedom to leave, without an exit fee
  • Ownership of everything you put on our hardware

GET STARTED

Ready to see it running on your own workload?

Book a 30-minute call with our team, or get a written quote for reserved or dedicated capacity within one working day.

Book a demo

Questions

About HuneX

HuneX is Danish-owned and Danish-operated. There is no US parent and no foreign holding structure, which is precisely why the US CLOUD Act does not reach the workloads we host.
No. We own and operate the hardware ourselves - power, cooling, networking and GPUs - in our own data center in Denmark. We are not aggregating spare capacity from third-party hosts the way peer-to-peer marketplaces do, which is why we can put a name and an address behind the uptime commitment.
You can, and for experiments it often makes sense. What those platforms do not give you is a named jurisdiction, a guaranteed uptime commitment, support in your own time zone, or a documented answer to where your data physically sits. When a compliance officer asks those questions, "it depends on the host" is not an answer.
On-demand capacity is billed by the hour with a minimum term of 5 year contract.