AI Lab

D
H
M
S
CONVOCATORIA CERRADA

Feasibility of training and inferring advanced large language models (LLMs) in data centers in Mexico and Brazil.

By: Tatiana Sandoval
Tatiana Sandoval

Latin America faces a growing challenge from the rapid expansion of AI, which demands enormous computational and energy resources, exacerbating its historical dependence on foreign infrastructures and limiting its technological sovereignty. To reverse this, it is vital to strengthen computing governance and democratize AI through regional frameworks like the OAS’s MIGDIA and the “Declaration of Santiago,” which seek ethical and responsible policies adapted to the local context. Given the high financial and environmental costs of building proprietary data centers, a more viable option is to adopt and specialize pre-trained models (e.g., DeepSeek) in key areas such as health, thereby reducing investment and carbon footprint. In short, the path to technological independence in AI lies in sustainable infrastructures, appropriate governance frameworks, and adoption strategies that balance costs, benefits, and sustainability.