- PRICING
One price sheet. No egress line. No surprises on the invoice.
On-demand or reserved GPU capacity, billed by the hour, hosted in our own data center on Danish soil. What you see below is what lands on the invoice - data transfer, support, and jurisdiction are never a separate line item.
99.9
%
Uptime SLA, reserved capacity
1
hr
Minimum billing increment
DK
/EU
Jurisdiction, always
ON-DEMAND
Pay by the hour.
Built for inference and fine-tuning workloads first - not spot-priced training clusters.
Every instance is single-tenant, bare-metal-backed, and metered in Danish kroner, euro or dollars, your choice.
Frontier AI training · Reasoning-intensive inference
NVIDIA B300
251 GB VRAM · 1× GPU per instance
$
4.73
/ GPU-hour
≈ €4.09/hour · on-demand
- vCPU 32
- Memory 288 GB HBM3e
- RAM 251 GB RAM
Large-scale AI training · High-throughput inference
NVIDIA B200
141 GB VRAM · 1× GPU per instance
$
4.07
/ GPU-hour
≈ €3.52/hour · on-demand
- vCPU 28
- Memory 141 GB RAM
- RAM 283 GB RAM
Fine-tuning · High-throughput inference
A100 SXM
80 GB VRAM · 1x GPU per instance
$
0,95
/ GPU-hour
≈ €0.82/hour · on-demand
- vCPU 16
- Memory 80 GB VRAM
- RAM 125 GB
■
Prices shown ex. VAT on a 5 year contract, metered per second after the first minute.
GPU availability by region depends on current cluster capacity - reserved terms guarantee allocation, on-demand does not.
GPU availability by region depends on current cluster capacity - reserved terms guarantee allocation, on-demand does not.
Cost-efficient inference · Batch workloads
NVIDIA L40S
48 GB VRAM · 1× GPU per instance
$
0.76
/ GPU-hour
≈ €0.65/hour · on-demand
- vCPU 16
- Memory 48 GB VRAM
- RAM 94 GB RAM
MOST USED FOR INFERENCE
Larger models · low-latency inference
NVIDIA H100 NVL
94 GB VRAM · 1× GPU per instance
$
1.91
/ GPU-hour
≈ €1.64/hour · on-demand
- vCPU 16
- Memory 94 GB RAM
- RAM 94 GB RAM
Fine-tuning · High-throughput inference
H100 SXM
80 GB VRAM · 1x GPU per instance
$
2.09
/ GPU-hour
≈ €1.79/hour · on-demand
- vCPU 20
- Memory 80 GB VRAM
- RAM 125 GB
■
Prices shown ex. VAT on a 5 year contract, metered per second after the first minute.
GPU availability by region depends on current cluster capacity - reserved terms guarantee allocation, on-demand does not.
GPU availability by region depends on current cluster capacity - reserved terms guarantee allocation, on-demand does not.
RESERVED CAPACITY
Commit to a term. We commit to the discount - and the GPUs.
Reserved capacity is the same hardware and the same contract terms as on-demand, at a lower effective hourly rate, with allocation guaranteed for the length of the term.
- → Switch between L40S and H100 within a reserved term at no penalty, capacity permitting.
- → Invoiced monthly in DKK or EUR - no upfront lump sum required for 1- and 3-month terms.
- → 12-month terms include a named account contact and priority capacity during regional demand spikes.
| Term | Discount | Effective rate |
|---|---|---|
| On-demand | — | €2.40 / hr |
| 1-month reserved | −15% | €2.04 / hr |
| 3-month reserved | −24% | €1.82 / hr |
| 12-month reserved | −35% | €1.56 / hr |
ALWAYS INCLUDED
The parts other providers meter separately.
// 01
No egress fees
Move data in and out as needed. We don't charge for bandwidth, and we don't lock your data in with punitive transfer pricing.
// 02
GDPR by design
Data processing agreements, EU-only sub-processors, and workloads that never leave Danish jurisdiction — not a compliance add-on.
// 03
Danish support
Reach an engineer in Danish or English, based in the same time zone as your team — not a follow-the-sun ticket queue.
// 04
Transparent metering
Per-second billing after the first minute, live usage dashboards, and one invoice — no separate line items for storage, IPs, or support tiers.
HOW WE COMPARE
Priced between the discount marketplaces and the hyperscalers - deliberately.
We won't win a race to the bottom against global GPU marketplaces, and we're not trying to. The market has room for a provider that trades the very lowest sticker price for guaranteed jurisdiction, predictable uptime, and a Danish contract.
| Discount marketplaces | Global neoclouds | Hyperscalers | HuneX | |
|---|---|---|---|---|
| Indicative H100/hr | ~€0.75–2.30, variable per host | ~€1.90–5.60 | ~€2.80–6.50 | €2.40, fixed |
| Uptime guarantee | None on shared/community tier | SLA on enterprise contracts | Strong SLA, high price | 99.9% on reserved terms |
| Data jurisdiction | Global, host-dependent | Often US-owned entity | Subject to US CLOUD Act | Danish soil, Danish entity |
| Support | Community / ticket only | Enterprise tier only | Enterprise tier only | Danish engineer, all tiers |
| Best fit | Non-sensitive, best-effort jobs | Large training clusters | Teams locked to the ecosystem | Compliance-sensitive inference & fine-tuning |
Figures are indicative market ranges, not live third-party quotes - request a tailored quote for your workload.
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.
Get in Touch
Talk to an Expert
Submitting this form is completely non-binding. One of our specialists will contact you shortly to discuss your needs.
PRICING FAQ
Questions procurement usually asks.
Yes. Billing is available in DKK, USD or EUR, and institutional customers can be invoiced monthly on standard net-30 terms rather than metered card billing.
Overflow usage is billed at the standard on-demand rate for the GPU type in use, on the same invoice - no penalty rate, no service interruption.
Institutional and multi-year agreements are quoted individually, and typically include a dedicated Data Processing Agreement, a named technical contact, and priority allocation. Contact sales for a tailored proposal.
All compute and storage runs in our own data center on Danish soil. Data does not leave Danish/EU jurisdiction as part of normal operation, and this is documented in our standard DPA.