Connect to European AI Infrastructure in Less Than 10 Minutes.
One place for finding and renting AI-ready servers in European datacenters. No developer experience needed: start a GPU server from the platform manager, or let your AI agent do the work for you.
- from V100 to B300 and GB300
- 10+ GPU types
- from sign-up to a running GPU
- < 10 min
- AI-ready servers, billed in euros
- from €0.15 / h
- platform manager, API, SDK, agent skill
- 4 ways in
Europe Has the Capacity. But No Single Way In.
AI-ready servers stand ready all over Europe. But they are scattered across fragmented providers and datacenters, and each operator is its own island: its own contracts, its own tooling, its own access process. Buying from one locks you to one.
Mimiry changes the landscape. We give you one control plane from which you can connect to servers across multiple different providers, giving you the best chance of finding scarce GPUs, with one way of connecting to all of them. Making it easier than ever for solo developers or teams to access compliant and secure AI-ready servers.
Think of Mimiry as one front door to Europe's AI infrastructure. You open one account and get one bill, and from there you can reach GPU servers at several providers, the same way every time. You choose what you need. Mimiry deals with the datacenters.
GPUs Available in European Datacenters.
Log in to the platform manager, or use the Mimiry availability endpoint, to check live availability.
B300
268 GB per GPUSXMB200
180 GB per GPUSXMH200
141 GB per GPUSXMH100
80 GB per GPUSXMA100 80GB
80 GB per GPUSXMA100 40GB
40 GB per GPUSXM or PCIeRTX PRO 6000
96 GB per GPUPCIeRTX 6000 Ada
48 GB per GPUPCIeL40S
48 GB per GPUPCIeTesla V100
16 or 32 GB per GPUSXM or PCIeYou Don't Need to Be a Developer.
Use Mimiry the way that suits you. Every way in reaches the same GPUs, the same account and the same bill.
Platform manager
Sign up, add your team, follow costs, and start or stop GPUs from your browser.
Open the platform managerAI agent skill
Tell your AI agent what you need, for example “train this model on an H100 and send me the result”, and it uses Mimiry for you.
Get the agent skillPython SDK
Put one line above a Python function and it runs on a European GPU. Install it with pip install mimiry.
Read the quickstartAPI
Connect Mimiry to your own tools and pipelines. The same GPUs, the same account, the same bill.
Start with the quickstartSecure and Compliant: Your Data Stays Under European Law.
What changes when your AI runs on European infrastructure, through a European company.
European, all the way down
Your workloads run on physical machines in European datacenters, under European law, through Mimiry ApS, a Danish company. Your contract and your invoice are European. Access works without passwords: every call is signed with your SSH key, so there is no password to phish, and your team gets its own users and roles in your organization.
Pay for the minutes you use, in euros
Every GPU has an hourly rate in euros that you see before you launch. You pay for the minutes your session runs, with a ten-minute minimum per session. No reserved-instance maths, no currency conversion on the invoice.
H100 80 GBEUR
fine_tune()2 h 14 min2 h 14 min
transcribe()47 min47 min
total3 h 1 min3 h 1 min
Nothing left behind
Each job runs in its own isolated environment, which is removed when the job ends. The data you want to keep lives on your organization's own storage volume, and nowhere else.
Mimiry ApS · Copenhagen, DenmarkEuropean datacenters · EUR
From First Experiment to Production Training.
Train and fine-tune models
Train on your own data, on the GPU size the job needs, and keep checkpoints on your own storage between runs.
Process large volumes
Run documents, audio or images through AI models in batches, and pay only for the minutes the job runs.
Find the right AI model
We test open models on your own examples and tell you which one works, on which GPU, at what monthly cost.
Talk to usFor your engineers
Four Primitives. That's the Whole SDK.
A decorator, an image builder, a bash escape hatch, and a volume you can keep. Provisioning, queueing, image pulls, and retries happen behind them, so you never think about them.
@mimiry.function
Decorate a Python function, call .remote(), get the result back as a Python object.
PYTHON
@mimiry.function(gpu="A100") def transcribe(url: str) -> str: ... transcribe.remote("https://…/audio.mp3")
mimiry.Image
Bring any container image and any pip or apt packages. Chainable, no Dockerfile required.
PYTHON
image = ( mimiry.Image .from_registry("nvcr.io/nvidia/pytorch:24.01-py3") .pip_install("torch", "transformers") .apt_install("ffmpeg") )
mimiry.run
Raw bash on a GPU when you don't need Python at all. For ffmpeg, llama.cpp, or any CLI tool.
PYTHON
result = mimiry.run( image="nvcr.io/nvidia/pytorch:24.01-py3", gpu="A100", command="nvidia-smi", )
mimiry.volume
A hundred gigabytes of NVMe that outlives the session. Create it once, attach it at launch, write to /data. The next run finds what the last one left.
PYTHON
@mimiry.function(gpu="A100", volume="ckpt") def train(step: int) -> None: torch.save(state, f"/data/ckpt-{step}.pt")
Install.
pip install mimiryRegister.
add your SSH key in the platform managerDecorate.
@mimiry.function, then call .remote()That's the whole onboarding. The same Python you already run, with a GPU under it.
One decorator between your laptop and an H100.
PYTHONembed.py
import mimiry image = ( mimiry.Image .from_registry("nvcr.io/nvidia/pytorch:24.01-py3") .pip_install("sentence-transformers") ) @mimiry.function(gpu="H100", image=image) def embed(text: str) -> list[float]: from sentence_transformers import SentenceTransformer model = SentenceTransformer("all-MiniLM-L6-v2") return model.encode(text).tolist() vec = embed.remote("hello from your GPU")
embed() on H100 80 GBQUEUED
Tell Us What You Want to Build.
Whether you need GPUs tomorrow or want to know which AI setup fits your company, write to us.
Or email info@mimiry.com