See all news
Image: El Economista
El Economista
Tech in Mexico & LatAm

Mexico has talent but few GPUs! Latin America proposes sharing AI infrastructure

Experts from ECLAC, the IDB and the tech industry said Latin America needs to share computing infrastructure and agree on common rules for data use to develop its own AI. Mexico holds about 2% of the region's identified GPUs.

On October 21, El Economista published an analysis of the proposals discussed during the presentation of the ILIA 2026. The study, prepared by ECLAC and Chile's National Center for Artificial Intelligence, identifies a strong concentration of computing resources: Brazil holds approximately 92% of the registered regional capacity, while Mexico accounts for about 2% of the identified GPUs. The figures correspond to the universe measured by the index, not to an absolute inventory of all the processors that exist in each country.

Representatives of industry and multilateral organizations proposed shared infrastructure networks, interoperability between providers and rules that make it easier to use data legally in research. The discussion included experiences such as LatamGPT and the difficulties of assembling training corpora with proper permissions. Mexico scored 52 points out of 100 and ranked eighth among 20 countries, despite having 466 companies linked to AI and a broad range of graduate education programs. The challenge pointed out is not just buying accelerators: it also involves coordinating universities, funding, licenses and projects that address regional productive needs.

At a glance

WhatRegional experts proposed sharing computing power and data under clear rules to reduce technological dependence and develop Latin American artificial intelligence.
WhenEl Economista published the information on October 21, 2026, based on the presentation of the Latin American Artificial Intelligence Index 2026.
WhereThe debate covers Latin America and the Caribbean; the report identifies a concentration of processing capacity in Brazil and relatively limited capacity in Mexico.
HowThe proposal combines shared computing centers, interoperability between providers, agreements on data licensing and scientific projects among countries.
WhyInfrastructure and data rights determine whether regional models can be trained and academic talent can be turned into useful products.

Build your digital presence

Start today: write to us and we’ll reply within 24 hours.