Quick Answer:
Chile's National Center for Artificial Intelligence (Cenia) is training Latam-GPT on a supercomputer at the University of Tarapacá. According to Euronews, it is being built collaboratively across eight countries with roughly a five million dollar supercomputer investment and more than eight terabytes of Spanish and Portuguese training data. It rewires the AI citation map for every LATAM business.
Key Takeaways:
- Supercomputer investment: approximately five million dollars to host the supercomputer at the University of Tarapacá in northern Chile, according to Euronews.
- Regional coalition: Euronews names eight participating countries — Chile, Uruguay, Brazil, Colombia, Mexico, Peru, Ecuador and Argentina — alongside universities, civil-society groups and government entities.
- Training corpus: Euronews reports the project has collected more than eight terabytes of Spanish and Portuguese training data; the first version was developed on Amazon Web Services.
- Prioritized use cases: Euronews highlights hospital logistical and medical resource management plus customer-service chatbots for airlines and retail, with Chilean company Digevo as an initial commercial partner.
- AEO implication: businesses with LocalBusiness schema, consistent NAP and Spanish-language regional FAQs are positioned to be cited when Latam-GPT enters production.
Something is shifting on the global AI map — and for the first time it is not shifting from San Francisco or Shenzhen. Chile, through its National Center for Artificial Intelligence (Cenia), has activated the supercomputer that will train Latam-GPT, the first large language model designed and built specifically for Latin America, in Spanish, Portuguese and the region's cultural contexts.
Whether your business operates in Santiago, Mexico City, Bogotá, Buenos Aires, Lima, Houston or Miami, this news is your concern. Latam-GPT is not a distant academic project — it is the infrastructure that will run the next wave of AI search across LATAM. And like every new citation layer, it will reward the businesses that prepare first.
What Latam-GPT Actually Is
Latam-GPT is a collaborative large language model (LLM), comparable in concept to GPT-4 (OpenAI), Claude (Anthropic) or Gemini (Google) — but trained on regional corpora instead of a predominantly English-language dataset. According to Euronews, the model is being developed by Cenia with the participation of eight countries: Chile, Uruguay, Brazil, Colombia, Mexico, Peru, Ecuador and Argentina.
Euronews also reports an initial development budget of roughly $550,000 — contributed primarily by the Development Bank of Latin America (CAF) — and that the first version of the model was trained on Amazon Web Services cloud infrastructure while the dedicated supercomputer was being installed. Cenia director Alvaro Soto told Euronews that models from other regions "include Latin American data in a reduced proportion."
The essential difference: Cenia's stated goal is not to compete on raw scale with English-trained frontier models, but to build a model whose dialect coverage and cultural references reflect Latin America. The gap shows up immediately in local queries.
The Tarapacá Supercomputer
According to Euronews, the supercomputer hosting the next training cycles is installed at the University of Tarapacá in northern Chile, with an installation cost of approximately five million dollars and a planned activation in the first half of 2026. The University of Chile is also part of the consortium. Euronews adds that the project has collected more than eight terabytes of training data.
The five-million-dollar figure is modest compared to the budgets of OpenAI or Anthropic, but it lands at a moment when the cost of training competitive regional-scale models has fallen sharply. The leverage in this project is not raw compute — it is the regional Spanish and Portuguese corpora that the coalition's universities, libraries and governments contribute.
Why Latam-GPT Is a Digital-Sovereignty Project
Cenia frames the initiative as critical infrastructure for the region's cultural and digital sovereignty. The logic is direct: if the models that respond to your citizens are trained on foreign data, the responses will reflect foreign biases. Product recommendations, legal norms, cultural references, idioms — everything is filtered through a lens that is not your own.
According to Euronews, early applications target hospital logistical and medical resource management and customer-service chatbots for airlines and retail. Euronews quotes Chilean President Gabriel Boric highlighting the region's "role in global technological development" and Minister Aldo Valle stating the tool aims to "break down prejudices." The Chilean company Digevo is named as an initial commercial partner developing specialized chatbots on the model. Cenia is providing Latam-GPT as a free platform.
What it means for LATAM:
- Government virtual assistants that understand local idioms without translation
- Education platforms adapted to national curricula, not generic templates
- More room for small businesses to be understood by AI in their own language
- Reduced dependence on foreign APIs with pricing and censorship policies alien to the region
What This Means for Your Business
Here is where MerchandisePROS asks you to pay attention. The activation of Latam-GPT is not an abstract technical headline — it is the announcement of a new AI citation layer that will progressively replace pieces of search behavior in LATAM. Just as Google AI Overviews now appears on more than 25% of US searches, Latam-GPT could become the preferred engine for regional queries.
The practical question for any business owner: when a consumer in Guadalajara, Medellín or Valparaíso asks "who is the best electrician near me?" or "where can I buy X?", will Latam-GPT cite your business or your competitor's?
The answer depends on the same five AEO signals we already recommend to US clients, now with a critical regional nuance:
1. LocalBusiness schema on your site — but with address, phone and hours in correct local format. Latam-GPT is being trained to distinguish "+56 9 1234 5678" from a phone string translated literally from English.
2. FAQs that mirror how customers ask in your country — a Mexican asks "¿cuánto cuesta?", an Argentine asks "¿qué precio tiene?", a Chilean asks "¿cuánto vale?". The regional model will recognize all three and cite whoever writes in those terms.
3. Authoritative citations on regional directories — not only Yelp and Google Business Profile but regional yellow pages, local chambers of commerce, MercadoLibre as a seller-presence verifier, OLX, and government business-registry initiatives.
4. Consistent NAP across Google, Facebook, Instagram and local directories — and consistent in regional format. A phone number "+52 33 1234 5678" must be exactly the same string everywhere.
5. A llms.txt file at the root of your domain — explicitly declaring who you are, what you do and where you operate, in Spanish. This is the emerging AEO signal almost no LATAM business is implementing today.
The Likely Timeline and How to Prepare
February 2026 coverage confirmed the initiative's official launch. May 2026 updates — including this report — focus on the supercomputer's activation and the first training milestones. Based on similar patterns (Bloom, Llama 2, European public-sector models), a reasonable projection is:
Q3 2026 — Q1 2027: first model checkpoints, academic evaluations, pilot deployments in public sector and universities.
Q2 — Q4 2027: integration with government platforms, ministerial virtual assistants in participating countries, first closed commercial pilots.
2028 onward: open API, integrations with regional search engines, possible appearance as the default answer layer in digital health portals, education platforms and government services.
That gives businesses 12 to 24 months to optimize their digital presence before the citation layer consolidates. A short window — but enough, if you act now.
Frequently Asked Questions
What is Latam-GPT?
Latam-GPT is a collaborative large language model (LLM) developed by Chile's National Center for Artificial Intelligence (Cenia) together with several Latin American countries. According to Euronews, it is trained specifically on Spanish, Portuguese and regional cultural contexts to counter the underrepresentation of Latin American data in leading models.
Where is Latam-GPT being trained?
The supercomputer dedicated to training Latam-GPT is installed at the University of Tarapacá in northern Chile. Euronews reports the installation represents an investment of approximately five million dollars, with activation planned for the first half of 2026.
Which countries participate in Latam-GPT?
According to Euronews, the participating countries include Chile, Uruguay, Brazil, Colombia, Mexico, Peru, Ecuador and Argentina. Participation combines universities, foundations, libraries, civil-society organizations and government entities.
When will Latam-GPT be available for commercial use?
The initial roadmap prioritizes public-sector and education use cases. An open commercial API is part of the later roadmap, with no public date confirmed at the time of this publication (May 2026). Businesses should prepare now to integrate Latam-GPT when it becomes available.
Why does Latam-GPT matter for my business?
Latam-GPT is being trained on Spanish and Portuguese regional corpora rather than English-translated data. This means that when it answers local questions about services, products and providers, it will be positioned to cite businesses with quality structured data and consistent NAP in those languages. Whoever optimizes their digital presence now will capture those citations before competitors react.
"Your next customer in LATAM is not going to ask Google. They will ask Latam-GPT. The question is whether your business exists for that model — or not."
- Diego Medina F, Founder of MerchandisePROS