AI Sovereignty: Can a Country Succeed Without Its Own Frontier Model?

National AI strategies increasingly emphasize domestic computing capacity, data governance and technological independence. The economic choices extend across the AI supply chain, from infrastructure and model development to public procurement and institutional capability.

AI sovereignty has become a central objective of national technology strategies. Governments are investing in computing infrastructure, supporting domestic AI companies and examining their dependence on foreign technology providers. Much of the discussion has concentrated on the development of national frontier models, particularly as the United States and China continue to dominate investment in advanced AI systems. Yet a country's ability to develop and govern AI depends on a wider range of capabilities, including access to computing resources, control over sensitive data, technical expertise and the terms under which public institutions acquire technology.

The OECD's 2026 report Securing AI Compute found that 44 of the 46 countries in the Global Partnership on AI, together with the European Union, had adopted national or regional AI strategies. Twenty-one jurisdictions treated computing infrastructure as a dedicated strategic priority, with specific policy instruments, capacity targets or budgets. This reflects growing recognition that access to computational resources influences domestic research, commercial development and the ability of public institutions to deploy AI. It also raises questions about how governments allocate capital across an increasingly complex technological supply chain.

What does AI sovereignty mean in practice?

AI sovereignty can refer to several distinct forms of national capability. A country may have the capacity to develop its own models, operate domestic computing infrastructure, govern the use of data, or maintain sufficient technical expertise to evaluate and adapt systems supplied by foreign companies. These capabilities can reinforce one another, although they require different investments and provide different degrees of independence.

The European Commission's 2026 approach to technological sovereignty emphasizes domestic innovation, diversified supply chains, control over critical data infrastructure and the ability to establish technical standards. Its strategy recognizes international partnerships as part of securing access to technology and reducing dependence on individual suppliers. This approach treats sovereignty as the capacity to manage strategic dependencies while participating in international markets.

A country may therefore exercise substantial control over AI deployment without developing a frontier model. Public institutions can operate specialized models on domestic infrastructure, maintain control over sensitive information and establish procurement requirements that permit systems to be audited, modified or replaced. Conversely, a domestically developed model may depend on imported semiconductors, foreign cloud infrastructure and software maintained by external providers. The location of model development offers only a partial account of technological dependence.

These distinctions are relevant to the economics of national AI strategies. Developing a competitive frontier model requires substantial expenditure on computational resources, specialist personnel, training data and repeated experimentation. Maintaining that capability involves continued investment as models and training methods improve. Countries must assess whether frontier development is necessary to achieve their strategic objectives or whether investment in infrastructure, applied research and specialized systems could support a broader domestic AI ecosystem.

Domestic compute and the economics of infrastructure

Computing capacity is increasingly treated as strategic infrastructure. Advanced AI systems require specialized processors, high-speed networking, substantial electricity supplies and facilities capable of managing demanding workloads. Governments can develop public computing facilities, support domestic providers or secure access through commercial and international arrangements. Each approach involves different financial commitments, operating responsibilities and levels of control.

Canada's sovereign AI compute strategy and the United Kingdom's Compute Roadmap illustrate the growing emphasis on national infrastructure. Canada has allocated funding for sovereign computing facilities and programmes that improve access for researchers and businesses. The United Kingdom has committed public resources to expanding advanced computing capacity and providing access through its AI Research Resource programme. Both approaches seek to develop the infrastructure available to domestic institutions, including those working on applications that do not require frontier-scale model training.

India has adopted a related approach through the IndiaAI Mission. According to government figures published in 2026, its shared computing capacity had expanded to more than 45,000 GPUs by June, with subsidized access supporting research and development projects. The programme also supports indigenous foundation models, including smaller language models and systems adapted to Indian languages. The combination of computing access and model development illustrates how public investment can support several layers of domestic capability.

Infrastructure ownership nevertheless carries substantial costs. Computing facilities require continued investment in hardware, energy, cooling, maintenance and security. The commercial value of the infrastructure depends on utilization, access arrangements and the needs of domestic researchers and businesses. For countries with limited financial resources, shared facilities, regional partnerships and negotiated access to external computing capacity may provide alternatives to fully domestic infrastructure.

Open-weight models and national capability

The availability of open-weight AI models has expanded the options available to governments and domestic technology companies. These models make their trained parameters accessible under specified licensing conditions, allowing developers to deploy and adapt them without relying exclusively on a commercial provider's application programming interface. They can be fine-tuned for local languages, public-sector procedures and specialized applications, subject to the technical capabilities of the model and the resources available.

For countries with smaller AI research sectors, open-weight models can provide a foundation for developing domestic applications and expertise. Universities, public agencies and local companies can adapt existing models to national requirements, develop evaluation methods and operate systems within approved infrastructure. The economic advantage comes from concentrating investment on adaptation, deployment and institutional capability rather than undertaking the full cost of frontier-model development.

Open-weight availability does not eliminate technical dependencies. Model licenses differ, training datasets may remain undisclosed, and deployment can depend on foreign hardware and software frameworks. Some models also require substantial computational resources. Governments therefore need to assess the practical conditions under which models can be operated, audited, adapted and maintained, alongside their suitability for particular public-sector applications.

The strategic value of open-weight systems is closely connected to domestic technical capacity. Access to model parameters becomes more useful when local institutions possess the expertise to evaluate performance, manage security, adapt models and maintain the surrounding infrastructure. Investment in research institutions, engineering skills and shared technical resources can strengthen these capabilities across public and private sectors.

Data governance and public procurement

National control over AI also depends on the legal and institutional arrangements governing data. Public administrations manage information concerning citizens, businesses, infrastructure and essential services. AI applications introduce questions about where that information is processed, which entities can access it and which legal jurisdictions apply. Domestic data storage may address some requirements, but effective governance also depends on contractual rights, access controls, security arrangements and the legal obligations of service providers.

The European Commission's Cloud and AI Development Act initiative illustrates efforts to distinguish levels of sovereignty assurance. Its framework addresses factors including infrastructure location, provider independence, ownership and control over software supply chains. The approach reflects the different requirements that public institutions may have depending on the sensitivity of the systems and information involved.

Public procurement is another instrument through which governments can influence technological dependence. Contracts can establish requirements for data portability, interoperability, audit access, security, continuity of service and the ability to change suppliers. These conditions affect the long-term costs of AI deployment and the extent to which public institutions retain control over essential functions.

The OECD's 2026 Digital Government Outlook found that only 21 of 36 surveyed countries provided central support for procuring AI goods and services, despite widespread public investment in AI initiatives. Procurement capability therefore represents a practical area of institutional development. Governments require sufficient technical and commercial expertise to assess suppliers, evaluate proposed systems and negotiate arrangements that remain appropriate as technology changes.

Procurement policy can also influence domestic industry. Public demand for specialized AI applications, local-language services and secure infrastructure can create commercial opportunities for domestic firms. The effects depend on the design of procurement processes, the competitiveness of local suppliers and the ability of public institutions to evaluate the products they purchase.

Sovereignty as an investment strategy

National AI strategies involve choices about where public investment can develop durable economic and institutional capabilities. Frontier-model development may be appropriate for countries with substantial research capacity, access to advanced computing infrastructure and the resources to sustain repeated development cycles. Other countries may prioritize shared computing facilities, specialized models, public-sector applications, data infrastructure or domestic engineering expertise. These investments can support participation in the AI economy at different points in the supply chain.

The European Union is pursuing several approaches simultaneously. Its AI Factories programme expands access to computing resources for researchers and companies, while the AI Gigafactories initiative is intended to support larger-scale model development and advanced industrial applications. These programmes address different requirements within the European AI ecosystem, from access to infrastructure to the development of advanced technical capabilities.

A useful assessment of national AI sovereignty would examine the degree of control a country exercises over essential systems, the availability of alternative suppliers, the capacity of domestic institutions to maintain and adapt technology, and the economic sustainability of its investments. It would also distinguish between capabilities that require domestic ownership and those that can be secured through partnerships or contractual arrangements.

Developing a frontier model is one possible component of this strategy. The wider objective is to establish the infrastructure, expertise, governance arrangements and industrial capacity necessary to develop and use AI in accordance with national priorities. The resulting degree of sovereignty will depend on the capabilities countries can sustain and the dependencies they are equipped to manage.

Sources and fact-checking

The article draws on recent institutional publications, including:

• OECD, Securing AI Compute (2026) for national strategies, infrastructure investment and computing capacity.

  OECD

  
  

• European Commission, technological sovereignty strategy (2026) for the distinction between domestic capabilities and strategic dependencies.

  EUR-Lex

  
  

• European Commission, AI Gigafactories for European computing infrastructure investments.

  European Commission

  
  

• Government of India, IndiaAI Mission (2026) for computing capacity and indigenous model development.

  PIB

  
  

• OECD, Digital Government Outlook (2026) for public procurement and government AI capabilities.

  OECD