Giotto Launches a Sovereign AI Family Built to Run Anywhere, Not Just in US Clouds

Giotto Launches Sovereign AI Models for Public Infrastructure

Swiss AI lab Giotto launched a new family of sovereign AI models on October 27, 2026, called Giotto Solo and Giotto Alto, with a pitch that is less about benchmark scores and more about control. The models are designed to run on public cloud, dedicated cloud in Switzerland or Europe, and on-premises hardware, and Giotto says Alto is available free to everyone. The company frames advanced AI as public infrastructure: something people and organisations should be able to use on their own terms, not only through someone else's cloud.

That framing is the story. Most frontier models today arrive as an API hosted in the United States, with your data travelling to infrastructure you do not control and terms you cannot renegotiate. Giotto is betting that governments, hospitals, banks and companies in Europe want a real alternative: a capable model they can run where they choose, under rules they understand. Sovereignty here is not a slogan. It is a deployment choice, and Giotto built the models to make that choice practical.

What Solo and Alto Actually Offer

Giotto Alto is the larger model of the pair, and Giotto says it runs on eight B300 GPUs. Giotto Solo is pitched as the smartest model that runs on a single GPU, which matters far more than it sounds. A model that needs a whole cluster is a procurement project. A model that runs on one GPU is something a mid-size company, a university lab or a government department can actually try this quarter, on hardware they may already own.

Both models are reasoning models, built for tasks that need multi-step thinking rather than one-shot answers: analysis, planning, document work and agentic workflows. Giotto is also launching apps for iOS and Android, plus free anonymous chat, so individuals can try Alto without creating an account. That free tier is a deliberate contrast with the subscription walls around most frontier models, and it gives Giotto a way to build an audience before it asks anyone to pay for deployment.

Why Sovereign AI Models Matter Now

The timing is not accidental. Europe has spent the last two years writing AI rules, arguing about data residency, and watching its biggest companies build on American infrastructure. The EU AI Act is now in force, with high-risk deadlines shifted but prohibitions and transparency duties already live. Procurement teams are asking harder questions about where data goes, who can compel access to it, and what happens if a provider changes terms or cuts access.

Giotto's answer is architectural: no US dependency. Deployment through a European shared cloud, a dedicated cloud in Switzerland or Europe, or on-premises hardware means the customer picks the jurisdiction and the operator. For a Swiss bank, a German hospital or a French ministry, that is not a nice-to-have. It can be the difference between a project getting approved and a project dying in legal review. Giotto is not the first lab to make this pitch, but it is one of the few launching with both a free model and a single-GPU option on day one.

The Hard Questions Giotto Still Has to Answer

A launch announcement is not a track record. Giotto has not yet shown independent benchmark results that would let buyers compare Solo and Alto against the established frontier models on equal terms, and efficiency claims always need third-party testing before they should be taken at face value. The models also need an ecosystem: developer tools, integrations, support and a community that can answer questions at 2 AM when a deployment breaks.

There is also the question of scale. Running on eight GPUs is efficient, but the largest models from US and Chinese labs still lead on raw capability for the hardest tasks. Giotto's bet is that most real work does not need the largest model, it needs a good model in the right place with the right controls. That bet will be tested by the first wave of enterprise pilots, and by whether those pilots turn into renewals.

Who Should Pay Attention, and What to Watch

If you run IT or compliance for a European organisation, Giotto is worth a pilot, not a migration. Test Alto on your own workload, measure it against the model you use today, and check the data residency story against your actual legal requirements. If you are a developer, the single-GPU Solo is the interesting one: it lowers the cost of experimenting with reasoning models locally, without sending data anywhere.

Watch three things over the next few months. First, independent evaluations of Solo and Alto against comparable open and closed models. Second, named enterprise or government deployments, not just launch partners. Third, pricing for dedicated cloud and on-premises support, which is where sovereign AI projects usually succeed or stall. Giotto has made the case for control. Now it has to prove the models are good enough that control does not feel like a compromise. For context on the wider pricing war, see how OpenAI halved the price of its work model, and for the regulatory backdrop, see the UK AI biosecurity lab funding.