Latest

Artificial Intelligence

Latin America is building its own AI: these are the 11 sovereign AI projects

Latin America is making progress on sovereign AI with 11 projects in eight countries. Brazil, Chile, Mexico and other countries seek to reduce their dependence

Latin America is building its own AI: these are the 11 sovereign AI projects
Latin America is building its own AI: these are the 11 sovereign AI projects

Latin America is beginning to build its own artificial intelligence infrastructure. Supercomputers, data centers, research clusters and models tailored to local needs are part of a global race in which governments seek to reduce their technological dependence on the United States and China.

The Center for a New American Security (CNAS) published version 2.0 of its Sovereign AI Index, a global map that tracks 185 government initiatives aimed precisely at reducing dependence on a small group of technological powers in the development of artificial intelligence (AI).

The report, with data updated through June 2026, offers a revealing snapshot of how governments, including those in Latin America, are trying to gain room for maneuver vis-à-vis the United States and China, countries that hold 90% of the computing capacity needed to develop and deploy frontier AI models.

What sovereign AI is

There is no single, agreed-upon definition of sovereign AI. The CNAS index considers any government-backed initiative explicitly tied to national strategic interests and sustained through material public investment in computing infrastructure, models or data ecosystems to be a sovereign AI project.

The reasons driving these projects are diverse. They include national security, economic development, cultural and linguistic preservation, and the desire to gain autonomy from foreign technological dependence. Often, several of these objectives appear combined within a single initiative.

The report classifies projects into three broad categories that represent fundamental layers of the artificial intelligence value chain. The first is infrastructure, which includes data centers, supercomputers, GPU clusters and programs for access to computing capacity. It is by far the dominant category, accounting for 59% of the total.

The second corresponds to models, that is, government-backed efforts to develop or adapt foundation models. They account for 32%. The third is data: initiatives aimed at building national datasets. They make up just 9% of the projects identified.

Eleven initiatives in eight countries

Within this global landscape, Latin America appears as a region whose activity is still incipient but beginning to grow. The index counts 11 sovereign AI initiatives spread across eight Latin American countries. The figure is far below Europe, with 79 projects, and Asia-Pacific, with 54, but it places the region ahead of Eurasia, with eight; Africa, with seven; and Canada, with three.

The comparison reflects one of the report's main findings: global investment in sovereign AI remains heavily concentrated in the Middle East and East Asia. Together, both regions account for more than 80% of declared investment. Latin America, by contrast, is moving forward with considerably more modest resources and ambitions.

Among the Latin American projects identified by CNAS are:

  • Mexico-Coatlicue: supercomputing for science and government, at the announcement stage.
  • Dominican Republic-CEIA-RD: national AI center for computing and training, in development.
  • Colombia-IA Hecha País: models for biodiversity, climate and health, with one model in the pilot phase.
  • Brazil-SoberanIA: AI in Portuguese with domestic computing, with the model already operational.
  • Brazil-Santos Dumont upgrade: expansion of computing capacity for AI and scientific research, operational.
  • Brazil-Jairu: AI research cluster at the University of São Paulo, operational.
  • Chile-Latam-GPT: a regional model led by Chile and developed with contributors from several countries, whose first version has already been released.
  • Paraguay-Yguazú Digital: an AI computing center developed with Taiwan, announced.
  • Uruguay-Antel Pando expansion: AI computing capacity for government and businesses, in the process of being acquired.
  • Argentina-Clementina XXI + IPAC: a supercomputer and an access program for research, operational and counted as two initiatives that share the same system.

The case of Latam-GPT is especially illustrative. Although the project is led by Chile, it was conceived as a regional effort and has contributors from different countries. Its approach shows one of the possible paths for Latin America: given the enormous technological and economic scale that artificial intelligence requires, some countries may choose to pool efforts and build regional alliances instead of developing exclusively national infrastructures and models.

The index also makes clear that talking about technological sovereignty does not necessarily mean having real technological autonomy. According to CNAS, more than three-fifths of the projects tracked worldwide declare at least one foreign partner. And four out of five of those partners are US companies.

The NVIDIA case shows how far this dependence persists. Its GPUs are used in 45% of all infrastructure projects registered globally.

The dependence does not disappear at the model layer either. Among projects built on open-weight models, Meta's Llama family is by far the most used. It is followed by Mistral, from France; Google, from the United States; and Alibaba, from China. The report itself concludes that, even when governments try to build sovereign AI, their projects tend to rest on US technological foundations.

Here the great paradox of sovereignty in artificial intelligence appears: countries want to reduce their foreign dependence, but to do so they need to turn to technologies, chips, models and companies from precisely the technological ecosystems they are trying to depend on less.

For the authors of the report, no country—not even the United States—can exercise complete control over all the inputs needed to develop frontier AI. Even the most technologically advanced economies depend on chips and manufacturing equipment produced in other countries.

The question, therefore, changes. For most economies, the real question is not whether they can achieve absolute sovereignty in artificial intelligence, something the authors consider unviable, but how to manage their dependencies. That means deciding which layers of the technology chain are strategic and must remain under greater national control and which ones can acceptably continue to depend on foreign suppliers.

What it means for Latin America

For Latin America, the Sovereign AI Index allows for two complementary readings. The first is that the region has already begun to move. Brazil, Chile, Colombia, Mexico, Argentina, Uruguay, Paraguay and the Dominican Republic have concrete initiatives related to computing capacity, language models or both areas.

Many of them are also linked to specific needs of the countries and the region, from language preservation to public health, biodiversity or environmental management. The second reading has to do with scale. The volume and dimensions of these projects remain modest compared with those developed in the Middle East or East Asia. The report points to limitations related to financing, energy infrastructure and installed computing capacity.

The race for sovereign AI in Latin America, therefore, does not necessarily consist of reproducing the technological model of the United States or China on a smaller scale. The CNAS report sets out a more pragmatic scenario, where instead of aspiring to complete technological self-sufficiency, the challenge consists of designing partnerships, deploying technologies and building governance frameworks that allow countries to retain significant decision-making capacity over how artificial intelligence is used within their borders.

Alfredo Collosa

Author

Alfredo Collosa

Alfredo Collosa is a tax specialist. Public accountant. Bachelor's degree in Business Administration. Official Master's in Public Finance and Tax Administration from UNED/IEF in Spain. AMLCA Anti-Money Laundering Certificate (Florida University, US) and member of MetaLaw

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading