---
title: "Become an EU AI inference provider · GPU capacity partners"
description: "Run a data centre or an inference platform in Europe? Supply GPU capacity or model endpoints to companies that need AI processed in the EU."
canonical: "https://inference.eu/become-a-provider/"
language: "en"
---

For data centres and EU inference providers

# Supply the European capacity behind inference.eu

We sell EU-hosted AI to companies that can't send data outside Europe. If you run GPUs or serve models in the EU or EEA, we'd like to talk about putting your capacity behind it.

[Apply as a provider](https://inference.eu/contact/?interest=provider)[What we look for](#criteria)

1.  01Your hardware, your data centre
2.  02GPU capacity or model endpoints
3.  03EU or EEA locations only

## What working with us looks like

You keep running your infrastructure. We bring the customers, the API and the contracts on the customer side.

-   ### Demand from European companies

    Our customers choose us because their data has to stay in Europe. Providers supply the capacity those requests run on.

-   ### You keep your hardware

    GPUs stay in your data centre and are run by your team. We agree capacity, models and terms with you up front.

-   ### Capacity or a full endpoint

    Offer GPU capacity for models we deploy, or serve models yourself behind an OpenAI-compatible endpoint.

-   ### One clear set of data rules

    Providers process data in the EU or EEA and sign a data processing agreement, so customers see one clear subprocessor list.

## Who we're looking for

If most of these describe you, get in touch. If one doesn't, tell us anyway.

-   ### Data centres in the EU or EEA

    Every location that would handle customer requests is in the European Union or the European Economic Area.

-   ### Current-generation GPUs

    H100, H200, B200 or comparable accelerators, with the memory and networking to serve large open models.

-   ### A European company

    Your contracting entity is based in Europe, so customer data stays under European law end to end.

-   ### Open to a DPA and audits

    You can sign a data processing agreement and answer the security questions our customers ask us.

## How it works

1.  1

    ### Tell us about your capacity

    Locations, hardware, the models you serve today and roughly how much capacity you could offer.

2.  2

    ### We agree models and terms

    A call to match your capacity with customer demand and settle volume, pricing and the DPA.

3.  3

    ### Test traffic, then live

    We send test requests first. Once latency and reliability look right, real traffic follows.

## Common questions

The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly.

It depends on the setup. Reserved GPU capacity and per-token arrangements for model endpoints are both on the table, and we agree the model with each provider.

No. If you already serve models behind an OpenAI-compatible API, we can route to it. If you offer raw GPU capacity, we agree how models are deployed on it with your team.

Open-weight models our customers use, such as Llama, Qwen, DeepSeek, Mistral and GLM. We agree the list with you before any traffic arrives.

Locations in the EU or EEA are a requirement, because that is the promise we make to customers. A company from elsewhere with European data centres can still apply.

Customers see providers on the subprocessor list that comes with their DPA. Anything beyond that, such as case studies, only happens with your agreement.

[Ask a question](https://inference.eu/contact/)

## Put your European capacity to work

Send us your work email and a short note on your locations and hardware. We reply within a few working days.

-   Reply within one working day
-   Pricing for your volume
-   DPA ready to sign