Big AI models may be global by definition, but according to Gartner, a movement could take this expansion of AI use in a different direction. According to a recent study, by 2027 around 35% of countries will turn to region-specific AI platforms built on proprietary contextual data and aligned with local laws, languages and values.
Today, this rate is around 5%, which indicates a structural change in the way governments and companies are approaching the use of technology. The trend reflects a combination of geopolitical, regulatory and national security pressures.
Gartner’s view is that, in a scenario of growing distrust in relation to closed and concentrated models, countries with digital sovereignty agendas have started to invest in their own AI stacks, which include everything from computing power and data centers to models trained with local data.
Read more:
“Countries with digital sovereignty goals are increasing investments in national AI stacks as they seek alternatives to the closed US model,” says Gaurav Gupta, vice president analyst at Gartner. “Decision makers are prioritizing solutions that align with values, regulatory frameworks and local expectations, and not just those with the largest volumes of training data”, he assesses.
In practice, this has driven the development of localized models. According to Gartner, regional Large Language Models (LLMs) already outperform global ones in applications such as education, public services and legal compliance, especially in languages other than English. The gain in contextual value, in these cases, outweighs the scale advantage of generalist models.
Continues after advertising
This movement, however, comes at a high cost. Gartner estimates that countries that choose to establish a sovereign AI stack will need to invest at least 1% of Gross Domestic Product (GDP) in AI infrastructure by 2029. The figure reflects not only the cost of hardware and data centers, but also of development, maintenance, governance and continuous updating of models.
Read more:
AI sovereignty, in this context, goes beyond the technology itself. This is the ability of a nation or organization to control how AI is developed, implemented and used within its borders. Local regulations, data residency requirements, national security concerns and corporate risks have accelerated this movement.
Continues after advertising
“The fear of falling behind in the technological race also weighs heavily”, points out Gartner. Countries and companies are being pressured to innovate quickly to avoid excessive dependence on foreign suppliers and guarantee autonomy in strategic areas.
In Brazil, some deeptechs are already following this movement. Names like WideLabs are winning over customers with the proposal of offering a regionalized model. In fact, in conversation with the Startups Last year, CEO Nelson Leoni revealed that several Latin American countries were already interested in the startup’s model, as a regional and sovereign AI option.
Read more:
Continues after advertising
The Federal Government has also entered this race. In addition to launching the Brazilian Artificial Intelligence Plan last year, the Ministry of Management and Innovation in Public Services (MGI) announced last October that the federal administration has plans to spend around R$23 billion on sovereign AI projects over the next four years.
However, in this scenario, infrastructure becomes the central element. “Data centers and AI factories form the critical backbone of the stack that enables AI sovereignty,” says the Gartner analyst. In Brazil, this race has also begun: an example is the investment of R$3 billion that Scala Data Centers announced for the creation of “AI City” in Eldorado do Sul, in Rio Grande do Sul.
