The Lausanne-based artificial intelligence (AI) lab Giotto is launching Solo and Alto, two reasoning models developed in Switzerland and designed as sovereign AI. According to ICTjournal, which reported the announcement on 28 October 2026, they are meant to run on European infrastructure or directly on the user's premises. The start-up thus claims to offer an AI that does not depend on the American cloud.
What Giotto is announcing with Solo and Alto
The two models differ first and foremost in the hardware they require. Giotto Solo is designed to run on a single graphics processor (GPU, for graphics processing unit). Giotto Alto targets more demanding reasoning tasks and professional uses: it runs on eight Nvidia B300 GPUs.
This range builds on two earlier steps, according to ICTjournal. Giotto presented its "portable AI" approach in May, then opened a free cloud version of its AI in July.
Public access is broad. The launch includes a free and anonymous chat service, usable without creating an account, as well as apps for iOS and Android. Alto is made available free of charge on Giotto's shared cloud, and an application programming interface (API) is offered to developers.
How Giotto deploys its sovereign AI
The sovereignty Giotto claims rests on the choice of where the models run. According to ICTjournal, Solo and Alto can be used in three ways:
- through a shared cloud, hosted and operated in Europe;
- through a dedicated, isolated cloud, in partner data centres located in Switzerland or elsewhere in Europe;
- on premises, on compatible hardware, from a single-GPU workstation for Solo to multi-GPU server racks for Alto.
Giotto states that its cloud services are aligned with the requirements of the European Union's General Data Protection Regulation (GDPR). It also asserts that they do not depend on American cloud infrastructure. These are the company's own statements, which the article does not detail.
Aldo Podestà, CEO and co-founder of Giotto, sums up the lab's position in the press release: "Europe must offer more than another flag on the same dependency. Sovereignty must belong to the people and organisations that use AI, and not only to the companies that develop it."
What method lies behind these reasoning models?
Giotto is betting on "test-time compute", Aldo Podestà explained in an interview with the magazine Bilan. This approach allocates more computing power at the moment the model answers, rather than training ever-larger models.
The stated goal is efficiency. Giotto says it designed its models to reduce computing, energy and infrastructure needs. This is what would make it possible to run Solo on a single graphics card.
A distinction must be made here between fact and promise. The hardware configuration is an announced fact: one GPU for Solo, eight for Alto. The reduction in consumption remains a claim by the company, with no figure published in the sources consulted.
What has been measured, and what has not yet
The ICTjournal article cites no benchmark results. So neither the quality of Solo's and Alto's answers nor their ranking against other models is known.
Several pieces of information useful to a business are also missing: the size of the models, their training data, their licence, and the price of the API and the dedicated cloud. The partner data centres in Switzerland are not named.
Giotto versus Apertus and the American models
Giotto is not alone in championing AI designed in Switzerland. ETH Zurich and the Swiss Federal Institute of Technology in Lausanne (EPFL) have launched the Swiss AI Initiative. According to the NZZ, cited in our article on AI specialists in Switzerland, this initiative notably supports Apertus, an open AI model.
The two efforts are different in nature. Apertus comes from an initiative of the universities, Giotto from a start-up that also sells a service. None of the sources consulted compares their performance.
Against the American giants, the argument is one of control. Andreas Krause, director of the ETH AI Center, put it this way to the NZZ: "When our best minds work for American groups, control of the technology ultimately lies in the United States."
Other Swiss players are targeting the same demand. The IT group Elca offers a sovereign AI platform, where data and models remain under the client's control, as our article on Elca and Beyond Civic pointed out.
| Offering | Provider | Stated hardware | Where it runs, according to the sources |
|---|---|---|---|
| Giotto Solo | Giotto, Lausanne start-up | A single GPU | European cloud, dedicated cloud in Switzerland or Europe, or on premises |
| Giotto Alto | Giotto, Lausanne start-up | Eight Nvidia B300 GPUs | Same options; free on the shared cloud |
| Apertus | Swiss AI Initiative (ETH Zurich and EPFL) | Not specified | Not specified; open model according to the NZZ |
| Sovereign AI Platform | Elca, Swiss IT group | Not specified | Data and models under the client's control |
What this changes for a Swiss company
Where the data is processed is the first question for a Swiss company. In our guide to AI for SMEs, Cyril Louis, CEO of the consulting firm Mavericx, points out that an AI agent can depend on several external models and services. Not all of them are operated in Switzerland.
Swiss law applies whatever the model. The Federal Data Protection and Information Commissioner (FDPIC) pointed out on 24 September 2025 that the Federal Act on Data Protection (FADP) applies directly to AI. Our page on AI regulation in Switzerland details this framework.
Our analysis: Giotto invokes the European GDPR, not the FADP. A Swiss company will have to check this point for itself before entrusting personal data to the shared or dedicated cloud.
The on-premises option most directly addresses the question of data residency. It does, however, require hardware: a single-GPU workstation for Solo, servers with eight Nvidia B300 GPUs for Alto. The sources do not put a figure on this investment.
Before any trial, a few precautions still apply:
- know where the data is processed, for what purposes it may be reused and who can access it, as recommended by the expert Alexandre Buga in the magazine PME;
- do not carelessly enter confidential information into an AI, as warned by the Federal Office for Cybersecurity (NCSC) on 16 October 2026;
- reserve the free and anonymous service for uses without sensitive data, as long as its terms are not known.
The demand exists among the public. According to the DigitalBarometer 2026, 76% of respondents put data protection ahead of convenience. A "Made in Switzerland" label strengthens the trust of 83% of them in a digital service, compared with 53% for a European label.
The legal framework could still change. According to the SME portal of the State Secretariat for Economic Affairs (SECO), Switzerland has signed the Council of Europe AI convention. A preliminary draft regulation is to be drawn up by early 2027.
What we don't know yet
- How Solo and Alto perform in recognised tests, and how they compare with Apertus or the American models.
- The actual computing and energy consumption, which Giotto says is reduced.
- The price of the API, the dedicated cloud and on-premises licences.
- The names of the partner data centres in Switzerland.
- The models' licence and the origin of the training data.
- Compliance with the FADP, beyond the claimed alignment with the GDPR.
Sources
- ICTjournal, Alexia Muanza, "Lausanne-based Giotto launches two AI models focused on sovereignty", 28 October 2026, based on Giotto's press release and an interview with Aldo Podestà in Bilan
- 2clics.com, articles cited in the text, based on the NZZ, the magazine PME, Inside IT, the FDPIC, the NCSC, the DigitalBarometer 2026 and SECO's SME portal
