D2.1. National AI Stakeholder and Market Landscape Report

Executive Summary Serbia’s artificial intelligence (AI) landscape is evolving rapidly. It is being shaped by parallel advances in digitalisation, innovation, industrial adoption and emerging regulatory frameworks. In recent years, the country has invested in modernising public services, strengthening its research and innovation capacity, and supporting the growth of high-tech companies and start-ups. These developments have created a foundation for the more systematic and large-scale deployment of AI across key sectors, including Health, Energy, Environment & Sustainability, and Culture & Language domains, where growing technological capabilities and a progressively more mature ecosystem can be developed. The primary objective of this report is to establish an empirical and analytical foundation for positioning Serbian Artificial Intelligence Factory Antenna (SAIFA) as the central national mechanism for integrating the Serbian AI ecosystem into the wider European high-performance computing (Europa) value chain. In strict alignment with the project proposal's description, the deliverable provides a detailed mapping of Serbian AI ecosystem actors, business domains, regulatory drivers, and barriers to adoption, alongside a dedicated focus on four priority verticals: Health, Energy, Environment & Sustainability, and Culture & Language. The report is structured to provide a comprehensive assessment of Serbia’s emerging AI ecosystem, combining desk-based analysis with planned field research. It first presents the methodology applied throughout the assessment, including the main analytical approaches, data sources and criteria used for mapping and evaluating the national AI landscape. The subsequent sections present the findings of the desktop research, including an assessment of the Serbian AI landscape and an analysis of stakeholders relevant to the SAIFA project. The AI landscape assessment covers four key dimensions of the national AI ecosystem: research & education, industry & commercial AI activities, regulatory framework, and computational infrastructure. The stakeholder analysis, in turn, identifies the organization’s and stakeholder groups relevant to SAIFA, examining their roles, capabilities, needs and potential contribution to the project and its activities. Starting in September 2026, the assessment will be complemented by field research involving interviews, surveys and focus groups with different stakeholder groups. This research will provide a more in-depth understanding of the barriers, needs, opportunities and potential for AI development and adoption in Serbia, complementing the evidence gathered through the desktop analysis. The findings from the field research will subsequently be consolidated and disseminated through dedicated thematic SAIFA White Papers, providing evidence-based insights into AI readiness, adoption barriers, user needs and development opportunities in selected areas. The results will also be formally incorporated into Deliverable D2.3, ensuring that the evidence collected through field research feeds directly into the subsequent stages of the SAIFA project. The entire research architecture is designed to remain coherent with international benchmarks and the engagement models of sister European AI Factories, specifically Pharos and IT4LIA, while ensuring strict compliance with General Data Protection Regulation (GDPR) and data management protocols. Alongside this report, a structured Serbian AI Ecosystem Database was developed, providing a tabular overview of AI-related scientific and research projects, publications by authors affiliated with Serbian institutions, patent applications, study programmes, AI companies, ecosystem enablers, support programmes, regulatory frameworks and regulatory-development initiatives, public institutions, and national AI and computing infrastructure. The database is designed to be made publicly accessible through the SAIFA website and to serve as a living resource that can be regularly updated as the Serbian AI ecosystem evolves. The assessment of macroeconomic and general digital preconditions positions Serbia as a fast-advancing adopter and the leading Western Balkan (WB) economy in AI readiness. This trajectory is underpinned by a highly competitive Information and Communications Technology (ICT) sector, which achieved a record export surplus and an all-time high total export value of €4.55 billion in 2025 (OITEG, 2026). However, a granular evaluation across the four conceptual pillars of the AI ecosystem reveals deep structural asymmetries: · Research, Innovation and Education: Serbia has introduced foundational AI components across primary and secondary curricula and accredited nine specialized AI-focused degree programs since 2021. Scientific publication output has experienced a sharp expansion, exceeding 450 AI-related papers in 2025. Conversely, the commercialization of this knowledge remains a critical bottleneck; only 66 AI-related patent applications were filed between 2017 and mid-2026, highlighting a weak technology-transfer pipeline from academia to market. · Industry and Firm-Level Intensity: The domestic AI business landscape comprises approximately 300 developers and integrators. Startups and Small and Medium-Sized Enterprises (SMEs) form the dominant core, accounting for 89% of these firms, showcasing high agility and an entrepreneurial drive that culminated in the emergence of Serbia’s first tech unicorn (Ominimo) in 2026. However, the corporate and venture capital layers are thin. Venture capital (VC) expenditure stands at a critical low of 8.0% of the EU average, leaving the startup base heavily reliant on public grants and personal funds, which severely restricts their ability to scale commercially. · Regulation and Governance: Serbia has actively established strategic policy intent, moving from its initial Strategy for the Development of Artificial Intelligence in the Republic of Serbia (SDAIRS) (2020–2025) and Ethical Guidelines to a comprehensive new SDAIRS for the Period 2025–2030. In the absence of specialized legislation, AI is currently governed by a distributed matrix of horizontal laws (data protection, cybersecurity, non-discrimination). The upcoming draft Law on Artificial Intelligence, largely aligned with the EU AI Act, signals Serbia’s move towards a more comprehensive AI regulatory framework. · Computational Infrastructure: Serbia’s computational infrastructure is developing rapidly, with significant investments in national AI and High-Performance Computing (HPC) capacity. Hosted at the Government Data Center in Kragujevac (GDCK), the National AI Platform has reached a combined operational capacity of 37 PFLOPS using NVIDIA A100 and H200 GPU architectures. The planned delivery of a third supercomputer based on NVIDIA Grace Hopper technology in early 2027 will further strengthen Serbia’s national AI and computing capacity. This report also provides a systematic Stakeholder Analysis, categorising ecosystem actors according to their functional roles within the AI ecosystem. The analysis distinguishes four main groups: Data Providers and Domain Experts (public institutions, utilities and industrial organisations holding valuable datasets and domain knowledge), AI Researchers and Model Developers (universities and research institutes providing methodological and machine-learning expertise), AI Application Developers (startups and software companies turning AI capabilities into applications), and End-Users and Beneficiaries (e.g. patients, farmers, businesses and public administrations whose needs define the objectives of AI solutions). The analysis is complemented by cross-cutting ecosystem enablers and intermediaries, including science and technology parks, industry associations, chambers, incubators and accelerators, as well as investors, funding organisations and public authorities. Their role is particularly important for matchmaking, dissemination, funding, regulation and creating the conditions for collaboration. Sector-specific analyses were conducted for healthcare, energy, environment & sustainability, and language & culture, identifying relevant stakeholder groups, concrete organisations, and the resources they can contribute or require. Illustrative use cases, presented in this section, demonstrate how organisations with complementary data, domain expertise, AI capabilities and market access can be connected through SAIFA. The analysis therefore proposes stakeholder engagement activities focused on turning connections into concrete projects, including thematic workshops and roundtables, AI-focused forums and hackathons, direct outreach and bilateral meetings, SAIFA information sessions, and participation in sector-specific and scientific events. Together, these activities position SAIFA not only as an AI and HPC service provider, but as a facilitator of collaboration, resource exchange and joint solution development across Serbia’s AI ecosystem. Serbia’s strongest AI potential lies in HealthTech and BioTech, as well as in environment and sustainability, where established research capacities, relevant data, industry demand and an emerging base of technology providers provide a solid foundation for further development. At the same time, culture & language, as well as cybersecurity, represent emerging high-potential areas, characterised by a growing number of AI-related projects, research activities and companies, although their ecosystem bases are currently less developed. These areas therefore offer significant opportunities for targeted ecosystem-building and are particularly relevant to SAIFA’s project priorities. The ecosystem is geographically concentrated around Belgrade, Novi Sad, Niš and Kragujevac, with major Research and Development (R&D) and computing capacities centred on leading universities, research institutes and national infrastructure. Acros