Switzerland Has Its Own AI Models. Who Will Buy Them?

Lausanne-based Giotto has introduced two AI reasoning models that businesses can run on European cloud infrastructure or their own servers. Winning customers from established international suppliers will depend on performance, costs and the work involved in switching.

Lausanne-based Giotto entered the enterprise AI market on 27 October with two reasoning models and a choice of hosting arrangements. Businesses can access the technology through a European cloud service, rent dedicated computing capacity or install it on their own servers. The Swiss company is competing with international developers whose models are already embedded in software used by businesses around the world.

Its smaller model, Giotto Solo, runs on a single graphics processing unit (GPU). The more powerful Giotto Alto requires eight Nvidia B300 GPUs, reflecting the computing demands of the tasks it is designed to handle. Both sit within a product range that Giotto wants to sell to organisations seeking greater control over their AI systems.

The company is also offering Alto through a free chat service and mobile applications. Those products allow individuals to try the technology without making a financial commitment, while the enterprise offering involves infrastructure, licences and technical support.

Giotto’s launch comes as Swiss and European businesses examine where their AI systems run and who controls the information they process. Demand for local hosting gives domestic developers an opening, although procurement teams still need to establish whether an alternative performs well enough to replace services they already use.

The Cost of Running AI

Giotto was founded in Lausanne in 2017 by graduates of EPFL’s mathematics programme. Its research has concentrated on machine learning and reasoning, including ways to reduce the computing resources needed to run AI models.

Solo’s ability to operate on one GPU brings the technology within reach of organisations that lack large computing facilities. Alto serves customers with greater processing requirements, either through dedicated cloud infrastructure or installations managed by the customer.

Running a model privately entails more than purchasing the necessary hardware. Businesses need capacity for employees using the system simultaneously, along with maintenance, security updates and technical staff to keep it operating.

Giotto offers several commercial arrangements. Its shared-cloud API includes a flat-fee service with limits on simultaneous requests, while customers needing additional capacity can purchase token packages. Dedicated installations carry monthly rental charges, and on-premises licences depend on the hardware involved.

A company with relatively predictable demand may find a dedicated installation easier to budget for than a service charged according to consumption. Another may prefer to buy computing capacity as needed rather than maintain equipment that remains idle for part of the day.

The economics depend on how employees use the model, the computing resources required and the assistance needed to integrate it into existing applications. Those questions enter procurement discussions alongside the model’s technical performance.

Benchmarks Are Only the Beginning

Giotto has published comparisons with international AI models and points to its second-place finish in the 2025 ARC-AGI competition as evidence of its research capabilities. The new model comparisons come from Giotto itself. Independent testing of Solo and Alto across a broad range of corporate applications is still needed to establish how they perform under comparable conditions.

Businesses buying AI have requirements that extend beyond published benchmarks. A financial institution processing internal documents needs the model to retrieve information accurately and respect access restrictions. Software developers assessing an assistant for coding work examine the quality of the generated code, its reliability and how well it fits into their development systems.

Testing becomes more demanding when a company already uses another model. Its applications, data connections and internal procedures may have been built around the existing supplier’s technology. Replacing that model requires developers to check the outputs again and establish whether the alternative works with the surrounding software.

Giotto’s free service gives potential customers an opportunity to examine its capabilities before committing to a larger installation. Moving from that initial trial to a company-wide deployment involves technical work that differs considerably between organisations. A strong benchmark result can attract interest, but procurement teams still need evidence from tasks that resemble the work their employees perform.

What Swiss Hosting Actually Provides

Giotto is making the location of its computing infrastructure a prominent part of the offer. Its shared-cloud service operates in Europe, with dedicated hosting available in Switzerland or other European data centres. Customers can also arrange installations on their own premises.

Financial institutions and other businesses handling confidential information have reasons to examine these options. Hosting affects where requests are processed, which organisations operate the infrastructure and how much control the customer has over access.

The contract determines the details. A prospective customer needs to establish whether requests are stored, how logs are handled and which suppliers take part in operating the service. An installation in Switzerland does not, on its own, answer every question about security or compliance.

Dedicated systems also require maintenance. Customers operating AI on their own hardware need arrangements for software updates, monitoring and technical problems, even when external cloud infrastructure is no longer involved.

Giotto says its European cloud service does not depend on American cloud infrastructure. That gives customers a specific claim to assess through technical documentation and contractual arrangements.

The supplier’s location may satisfy one procurement requirement while leaving performance and integration questions unresolved. Established AI providers have spent years developing tools that connect their models to business applications, making it relatively straightforward for existing customers to add new functions without changing supplier.

The Difficulty of Winning Enterprise Customers

Large technology groups already distribute AI through cloud services, office software and enterprise applications. Many businesses therefore gain access to new models through systems they have purchased for other purposes. A specialist developer has to give them a reason to introduce another provider. Some customers are seeking greater control over infrastructure or want to run models in environments they manage themselves. Others need to reduce operating costs or use technology suited to a defined business task. Those requirements vary enough that Giotto’s smaller and larger models will face different comparisons.

The purchasing process also involves people beyond the team interested in the AI application. IT departments assess integration and maintenance, security teams examine data handling, and finance departments compare costs. Employees testing the model need consistent results from their own documents and systems.

A new provider without an established enterprise customer base has limited deployment history to draw upon. Early installations will give Giotto opportunities to demonstrate how its models perform outside controlled tests, including the work required to maintain them.

Exporis works with technology companies on product positioning and B2B communications, where technical claims have to be translated into material that procurement teams can assess. In enterprise AI, that means explaining computing requirements, supported integrations, security arrangements and the results achieved in customer environments. Published benchmarks cover only part of the purchasing decision.

Giotto has chosen to compete through model development, flexible deployment and Swiss and European infrastructure. Businesses now have to decide whether those capabilities justify the cost and work of adopting another AI supplier.

A new AI provider has several audiences to convince before a contract is signed. Developers want documentation and examples they can test. IT managers need integration and maintenance details, while procurement teams examine licensing, operating costs and supplier responsibilities. These questions rarely fit into one product presentation. Technical comparisons, deployment guides and documented customer installations serve different stages of the purchasing process. A benchmark attracts attention; a completed implementation gives a prospective customer something closer to its own business to examine. Exporis develops specialist content and product communications around that progression, helping technology companies explain their products to the people responsible for buying and operating them.