BespokeYour own server

Your AI, on your own server.

Would you rather keep AI in-house? Together we choose the hardware suited to your models and volumes; we order it, install it on your premises or in the data centre of your choice, then commission and maintain it.

Bespoke pricing, depending on the hardware and services chosen.

Request a quote

The hardware

From the data centre to the office.

  1. NVIDIA Blackwell servers

    B200, B300 or RTX PRO 6000, running vLLM or SGLang: up to several hundred concurrent users, depending on the model.

  2. Mac Studio

    One or more, running llama.cpp or MLX, for a small team: quiet enough to sit in an office.

Our services

From hardware selection to maintenance.

  1. Selection

    We start from your use cases, your models and your volumes to choose the hardware that meets them, without oversizing it.

  2. Ordering and installation

    We order the hardware and install it on site.

  3. Commissioning

    We install the models and the inference engine, tune them for your hardware and open access to your teams and your applications.

  4. Maintenance

    We keep the models and the software up to date, and monitor that the server runs properly.

What runs on it

The whole SovIA range, in-house.

Your server runs what we serve from ours: SovIA, the legal assistant, sovereign inference, the legal API and, soon, agentic automation. Your data stays on your server, without any external AI provider.

Bespoke pricing: together we define the hardware, the services and the contract suited to your project, then send you a quote.

Request a quote