Comparison · breadth vs depth
Scaleway is a full European cloud — compute, storage and managed services at hyperscaler-style breadth; AxForge is narrow and deep — open-model inference on EU hardware it owns, and nothing else.
At a glance
| Dimension | AxForge | Scaleway |
|---|---|---|
| What it is | A narrow EU inference operator: a hosted open-model API plus dedicated machines, on hardware AxForge owns. | A full European cloud provider: compute, storage, networking and managed services. |
| Scope | One job — open-model inference. No general compute, no storage products. | Hyperscaler-style breadth: run an entire stack in one account. |
| GPU compute | Named machines rented monthly: NVIDIA DGX Spark (GB10, 128 GB unified memory) available now; RTX tiers on the way. | A range of GPU instances, rentable self-serve from the console. |
| EU footprint | EU-only by construction: eu-se-1 Stockholm and eu-es-1 Málaga live, more EU regions in deployment; region pinned per key. | A European company operating EU datacenters. |
| Model API | OpenAI-compatible, region-pinned, with zero prompt retention — Qwen3.8 27B served now, plus embeddings, image and speech. | Managed AI services offered as part of a broad cloud portfolio. |
| Hardware transparency | Down to the machine: named systems, published serving config, performance measured on the operator's own production node. | Standard cloud instance types — you choose a type, the platform runs the fleet. |
| How you buy | Waiting-list API keys and engineer-led onboarding; dedicated machines by monthly rental. | Self-serve sign-up with pay-as-you-go cloud billing. |
Honest routing
A full European cloud wins on these requirements:
| The whole stack | You want compute, storage, databases, networking and AI under one account with one provider. |
|---|---|
| Self-serve everything | You want to sign up today and provision instances, buckets and GPU capacity from a console without talking to anyone. |
| Managed services | You want the provider to run your databases, Kubernetes and object storage, not just your inference. |
| Large-scale GPU instances | You need big GPU instances inside the same cloud where the rest of your infrastructure already lives. |
Honest routing
| Transparent inference infrastructure | You want to know exactly which machine serves your tokens — named system, published serving details, numbers measured on that hardware. |
|---|---|
| GDPR-sensitive API workloads | GDPR compliant infrastructure: zero prompt retention, region pinned per key, TLS terminated in the EU. |
| Single-purpose focus | Inference is the product, not one line in a catalogue — the operator's engineers run the models and answer for them. |
| A dedicated machine | A DGX Spark serving only your traffic, as a flat monthly rental, available now. |
The substance
| Regions | eu-se-1 · Stockholm and eu-es-1 · Málaga, both live. More EU regions in deployment. |
|---|---|
| Data handling | Zero prompt retention — in-memory processing in Sweden, no disk, no logs, no training. Only billing metadata is kept. Full policy: axforge.ai/privacy. |
| API | OpenAI-compatible at https://api.axforge.ai/v1 — TLS 1.3, terminated in the EU. |
| Pricing | Qwen3.8 27B: €0.29 / 1M input · €1.77 / 1M output (launch pricing). Everything else is published too — see the price list. |
| Measured speed | 15.9 tokens/s single-stream decode for Qwen3.8 27B with speculative decoding. |
| Dedicated systems | NVIDIA DGX Spark (GB10, 128 GB unified memory), monthly rental, available now in Sweden. |
Speed measured on our production DGX Spark node, single-stream, 2026-08. We publish only numbers we measured ourselves.
FAQ
Only for AI inference. Scaleway is a full European cloud — compute, storage, networking and managed services. AxForge does one thing: open-model inference on owned EU hardware, as a hosted API and as dedicated machines. If you need a whole cloud, AxForge is not it; if you need transparent, single-purpose inference infrastructure, that is exactly what it is.
Depth on one layer: named machines rather than instance types, published serving details, performance numbers measured on the operator's own production hardware, and a model API with zero prompt retention. It is the difference between a catalogue and a workshop.
No. AxForge is inference-first infrastructure, not a general cloud. Most customers keep a general cloud for compute and storage and use AxForge for the inference path, where data sensitivity and auditability matter most.
eu-se-1 (Stockholm) and eu-es-1 (Málaga) are live,
with more EU regions in deployment. The region is pinned per API key with the
x-axforge-region header and echoed on every response.
Qwen3.8 27B is €0.29 per million input tokens and €1.77 per million output tokens (launch pricing), on an OpenAI-compatible API with zero prompt retention. Every other price is published too — see the full price list, or talk to an engineer.
Related