Service · Artificial intelligence

Cloud infrastructure for AI

Some organisations simply cannot let their documents travel to a shared service: a clinic holding patient histories, a law firm with case files, an accountancy practice keeping clients' books, a town council with residents' details or a manufacturer guarding its drawings. For them we build a sealed environment in the cloud, inside their own account, where the model runs and the data never leaves. We can use the regions the hyperscalers operate in Spain (AWS in Aragon, Microsoft Azure and Google Cloud in Madrid) or European providers such as OVHcloud, Hetzner, Arsys or IONOS. This is rented capacity: we sell neither servers nor graphics cards, and every bit of configuration happens remotely.

Region
Spain or the EU, as you prefer
GPUs
hired by the hour, never purchased
Search
over your own document base
Spend
alerts and timed shutdown

What falls within the scope of this service

Step one is separating running an existing model from tuning one on your data. The latter costs considerably more and is seldom required. These are the usual building blocks.

Pin down the details with one of our engineers

Region and provider

Spain if your policy or your customers insist on it, another EU region if it wins on price or GPU availability. We weigh AWS, Azure, Google Cloud and European providers against your actual case.

Compute capacity

GPU type and memory to suit the model size, and how you pay: hourly, with a usage commitment, or interruptible instances for overnight batch work.

Model in service

An open model deployed in your account, or the cloud provider's own managed offering in a European region, always on terms that forbid training on your content.

Questions over your documents

Storage, a vector database and indexing of protocols, contracts, manuals or data sheets, so the model answers while citing the internal source. Encrypted, versioned and with restore-tested backups.

Network and permissions

No public addresses, sign-in through Entra ID or another identity provider, permissions per department and every query logged for audit.

Spending kept in check

Dashboards showing GPU use, response times and running cost, alerts as the limit nears and automatic shutdown after hours.

Legal and sector requirements

Data flow documentation for your processing register and DPIA, health data as a special category under the GDPR, the National Security Framework (ENS) when you work for the public sector, and material so your lawyers can review what the European AI Act may demand.

How the engagement unfolds, one stage at a time

An existing model can be serving requests quickly. Tuning on your own data takes longer, depending on the amount of information and the GPU quotas the provider allows.

01

Requirements

Which task, what query volume, what kind of data and which location or sector constraints apply.

02

Design and cost

Architecture, recommended provider and monthly budget, with the simplest alternative always on the table.

03

Build

Infrastructure defined as code, model deployed and a secure link to your applications.

04

Operation

Training for your technical lead and full documentation, or ongoing care by us on a monthly contract.

What inflates the bill is not the GPU but the GPU left running with nobody using it. A test instance forgotten over a bank holiday weekend can cost more than a month of normal use. So from day one the environment has spend alerts, automatic shutdown out of hours and a monthly usage report that the person who signs the invoices can follow too.

Common questions

Not always. If your data is not particularly delicate and a commercial service contract guarantees processing in Europe with no training on your content, that route is usually cheaper and simpler. A private environment earns its keep when professional secrecy, health data or in-house know-how must stay firmly under your control.

Both satisfy the GDPR as far as location goes. Spain may be required by your clients, by a public tender or by your own policy. Other European regions sometimes have more GPUs free or keener prices. We settle it together after comparing the two.

We find out using your own documents and a question set drawn up with you. If you also work in Catalan, Galician or Basque, those go into the evaluation too. Public rankings reshuffle every few months; your results are what count.

That is why we describe the infrastructure as code and favour standard components. Moving is never free, but you are not tied to hand-made settings that only one person understands. And since we supply no hardware, there is no equipment to write off before you move.

That depends on the GPU type, the hours of actual use and how many documents are indexed. Before we start you get a monthly estimate covering two or three usage scenarios. Our set-up work is quoted per project or charged at €75 per hour + VAT.

Let us design your AI environment in the cloud

Describe the task and which data has to remain under your control. We will suggest an architecture in a Spanish or European region and a monthly budget.

Hours
Monday to Friday, 9:00-18:00 Spanish time (CET), answers within a working day
Meetings
Video calls via Google Meet or Teams

This site only stores the cookies it needs to work and to remember your chosen city. No advertising or tracking cookies are set. See our privacy policy for the details.