Skip to content
bona netto

02 · AI hosting

Compute next to the light.

GPU-ready servers sit directly on our fiber. Your data and your models are a few milliseconds apart, and nothing crosses the public internet to reach them.

One request, traced

Never leaves
our network.

From your office to the GPU and back again, every step happens on glass and hardware we run ourselves.

What you get

Built for
models.

Hosting planned around accelerators from the start: power, cooling and capacity in the same building as the fiber.

GPU servers

Discuss your model, memory and throughput requirements with an engineer to select the right hardware.

Private model hosting

Run open models or your own fine-tunes on hardware that is not shared.

Data stays here

Datasets, weights and answers stay on our network and in our racks.

Power and cooling

Racks sized for the heat and power draw of modern GPUs, not retrofitted.

Low latency

Hosting placed on the network to shorten the path between your data and your models.

Engineers on call

Agree the support coverage and escalation process that fits your workload.

Used for

What runs
on the light.

A few of the jobs AI hosting on our network is built for.

01

Private assistants

Language models for your own documents and staff, without sending data to a public service.

02

Training and fine-tuning

Move large datasets in fast over symmetric fiber and train next to them.

03

Inference APIs

Serve models to your apps with answers that come back in milliseconds.

04

Vision and media

Video and image models fed directly from cameras and storage on the same network.

Questions

Good to know.

Can I bring my own model?

Yes. Run open models, your own fine-tunes or anything that runs on the hardware you choose.

Is the hardware shared with other customers?

Dedicated servers are reserved for your workload. Ask us to specify the hardware and isolation in your proposal.

How does my office reach the servers?

Over a private link on our fiber, so the traffic never crosses the public internet.

How is it priced?

We quote for the hardware, capacity and support your workload needs. Talk to an engineer and we size it with you.

Availability

Does the light
reach you?

Tell us your business address. Our engineers will review connectivity options, capacity and the route to your building.