Silo AI

Silo AI is one of Europe's most significant AI companies, headquartered in Helsinki, Finland, and founded in 2017 with a mission to build practical AI solutions for enterprises. The company distinguished itself early on by focusing not just on research but on deploying AI in real-world business contexts across industries like manufacturing, automotive, maritime, and healthcare. Silo AI's client list includes major European and global enterprises such as Allianz, Philips, Rolls-Royce, and Unilever, demonstrating the breadth and maturity of its enterprise AI capabilities.

In July 2024, AMD announced the acquisition of Silo AI for approximately 665 million US dollars in an all-cash deal, making it the largest AI acquisition in European history at that time. The acquisition was completed in August 2024 and positioned Silo AI as a key component of AMD's global AI strategy. Despite the acquisition, Silo AI continues to operate from its Helsinki base with its established team and culture, now with the additional resources and hardware expertise that AMD brings to the table. This combination of European AI talent and world-class semiconductor capability creates a unique proposition in the global AI landscape.

The Poro Language Model Family

Poro, named after the Finnish word for reindeer, was Silo AI's first major open-source language model. It is a 34 billion parameter model specifically trained for English, Finnish, and code. Poro was trained using 512 AMD MI250X GPUs on the LUMI supercomputer located in Kajaani, Finland, which is one of the most powerful supercomputers in Europe and the world. The model was developed in collaboration with the University of Turku and represents a landmark achievement in building large language models that genuinely understand Finnish, a language that has traditionally been underserved by US-centric AI models due to its complex agglutinative grammar.

The significance of Poro extends beyond its technical specifications. It demonstrated that Europe could develop competitive large language models using European infrastructure and talent, challenging the notion that AI development was the exclusive domain of US tech giants. Poro is released under an open Apache 2.0 license, allowing organizations to download, modify, and deploy the model on their own infrastructure without licensing fees or restrictions.

The Viking Model Family

Viking represents the evolution of Silo AI's language model research, expanding beyond Finnish to cover all Nordic languages: Finnish, Swedish, Norwegian, Danish, Icelandic, and English, along with programming languages. The Viking model family comes in multiple sizes, including 7B, 13B, and 33B parameter variants, providing organizations with options to balance performance against computational requirements. Viking was trained on a massive dataset of two trillion tokens, making it one of the most comprehensive multilingual models focused on Nordic languages.

Viking 13B was developed using the LUMI supercomputer and an open-source training framework, which Silo AI has shared with the community to help other organizations replicate and build upon their work. The models demonstrate state-of-the-art performance on Nordic language benchmarks, significantly outperforming general-purpose models like GPT on tasks involving these languages. Viking is also available on Google Cloud, making it accessible for organizations that prefer cloud deployment over on-premises infrastructure.

Enterprise AI Solutions

Beyond language models, Silo AI provides comprehensive enterprise AI services that help organizations move from AI experimentation to production deployment. Their services include AI strategy consulting, where they help leadership teams identify the highest-value AI opportunities within their organizations, and custom model development, where their team of PhD researchers and engineers builds AI solutions tailored to specific business problems. Silo AI also offers MLOps implementation services, helping organizations establish the infrastructure and processes needed to deploy, monitor, and maintain AI models in production environments.

Silo AI's enterprise work spans several verticals. In manufacturing, they have developed computer vision systems for quality control and predictive maintenance models that reduce downtime. In the maritime industry, they have built AI solutions for route optimization and fuel efficiency. In healthcare, their work includes developing models for medical image analysis and clinical decision support. This breadth of domain expertise, combined with deep technical capability, sets Silo AI apart from more narrowly focused AI companies.

GDPR Compliance and European Data Sovereignty

As a Finnish company operating under EU jurisdiction, Silo AI is fully subject to the General Data Protection Regulation. This is particularly important in the AI space, where training data, model inference, and the handling of sensitive business information raise significant privacy concerns. Organizations working with Silo AI can be confident that their data is processed in accordance with European data protection standards, and the availability of open-source models means that sensitive data never needs to leave the organization's own infrastructure.

The European data sovereignty aspect of Silo AI's offering is especially relevant given growing concerns about dependence on US-based AI providers. With US AI companies subject to the CLOUD Act and other legislation that could compel data disclosure, European enterprises increasingly seek AI partners that can guarantee data remains within European jurisdiction. Silo AI's combination of European headquarters, LUMI supercomputer training, and open-source model licensing addresses this need comprehensively.

Training on LUMI Supercomputer

Silo AI's access to the LUMI supercomputer in Kajaani, Finland, has been instrumental in its ability to train competitive large language models. LUMI, funded by a consortium of European countries, is one of the world's most powerful supercomputers and represents a key piece of European AI infrastructure. The supercomputer is powered by AMD Instinct MI250X GPUs, which aligns naturally with Silo AI's now-parent company AMD. Training models on European infrastructure using European data is a powerful statement about the continent's ability to compete in the global AI race.

Open-Source Philosophy

Silo AI has committed to releasing its language models under open-source licenses, which represents a significant contribution to the European AI ecosystem. By making Poro and Viking freely available, Silo AI enables smaller companies, research institutions, and government organizations to access state-of-the-art Nordic language AI without the costs associated with proprietary API services. This open approach also fosters community development, as other researchers can build upon and improve the models. The decision to open-source is particularly impactful for smaller Nordic countries where the commercial incentive for large US companies to develop specialized language support is limited.

Competitive Positioning

In the European AI landscape, Silo AI occupies a unique position. While companies like Mistral AI focus on building general-purpose models to compete with OpenAI, and Aleph Alpha targets the German enterprise market with sovereignty-focused solutions, Silo AI combines Nordic language specialization with broad enterprise AI services. The AMD acquisition gives Silo AI access to cutting-edge hardware and a global sales network, while maintaining its European identity and Nordic focus. For enterprises operating in the Nordic region, Silo AI offers an unmatched combination of local language expertise, enterprise deployment experience, and open-source accessibility.

Limitations and Considerations

While Silo AI excels in the Nordic AI space, there are considerations for potential users. The enterprise consulting services are priced for medium to large organizations, which means smaller businesses may primarily benefit from the open-source models rather than custom engagements. The Poro and Viking models, while excellent for Nordic languages, are not designed to compete with the largest general-purpose models on English-language tasks or complex reasoning. Organizations seeking a direct replacement for ChatGPT or Claude for general-purpose use may find these models insufficient, though they are excellent for Nordic-specific applications.

The AMD acquisition also introduces questions about long-term strategic direction. While the acquisition brings significant resources, some stakeholders in the Nordic AI community have expressed concerns about whether the company's focus on open-source Nordic language models will continue as AMD integrates the team into its broader AI strategy. As of early 2026, however, Silo AI appears to be maintaining its commitment to open research and Nordic language excellence.

Who Should Use Silo AI?

Silo AI is ideal for Nordic enterprises that need AI solutions with genuine understanding of Finnish, Swedish, Norwegian, Danish, or Icelandic. Organizations in manufacturing, maritime, healthcare, and other verticals that want to deploy AI with European data sovereignty should consider Silo AI's enterprise services. Developers and researchers working on Nordic language applications will find the open-source Poro and Viking models invaluable, especially when data sensitivity requires on-premises deployment. For any organization that wants to leverage cutting-edge AI while keeping data within European jurisdiction and supporting the European AI ecosystem, Silo AI represents one of the strongest options available.