On request · not in the fleet
Not in the AxForge fleet today — because we buy hardware against registered demand. Tell us what you need, we build the business case together, and you hear first when the machines land.
How it works
You tell an engineer what you need — model sizes, traffic, term. When registered demand supports the purchase, we buy the hardware, and registrants hear from us first: first access, on machines sized to the workloads that asked for them. This is not theory — it is how the DGX Spark fleet grew to 4 live units with 7 more on the way. Register your demand and the next fleet gets built around it.
Specifications
| GPU | NVIDIA H100 |
|---|---|
| Memory | 80 GB HBM3 |
| Architecture | Hopper |
| In the AxForge fleet | No — this hardware is on request |
| Pricing | None exists — established with the business case |
Public NVIDIA specifications only. We publish performance numbers solely for hardware we run in production.
Available today
NVIDIA DGX Spark — 8 units available now in Sweden (eu-se-1) on monthly rental, 128 GB unified memory each, with measured inference numbers published on the GB10 page. RTX tiers are on the way and reservable.
FAQ
No. The H100 is not in the AxForge fleet today. We buy hardware against registered demand — tell us what you need and we build the business case with you.
No price exists — we do not own the hardware. Pricing would be established as part of the business case built with registered demand.
You tell an engineer what you need: model sizes, traffic, term. When registered demand supports the purchase, we buy the hardware and registrants hear from us first. That is how the DGX Spark fleet grew.
NVIDIA DGX Spark — 8 units available now in Sweden on monthly rental, with 3 more arriving in days. RTX tiers are on the way and reservable.
AxForge operates its own hardware in EU regions (Stockholm and Málaga live today). Any H100 capacity we deploy would follow the same model.