Where Latin America's cloud regions are and what they cost to use
Mexico and Chile have added regions from the major providers in recent years. Running close to users reduces latency and makes data residency easier, in exchange for shorter catalogues and, in some cases, higher prices.
Executive summary
Brazil has the longest-standing supply. AWS opened in São Paulo in 2011 and all four major providers now operate there. Between 2022 and 2025 Mexico gained five regions from AWS, Google, Microsoft and Oracle; Chile has four and expects AWS's by the end of 2026.
Colombia only has Oracle's region in Bogotá, and Venezuela and Panama have none. In those two countries the nearest region is abroad, so these providers cannot host data in-country.
The same virtual machine costs different amounts depending on the region. On AWS, a general-purpose instance costs almost the same in Mexico as in Virginia, and considerably more in São Paulo.
The local catalogue is shorter. AWS's Mexico region offers 285 instance types against 1,348 in Virginia, with no GPUs, and in São Paulo Google offers 4 of its generative AI models with a regional endpoint, and no third-party models, against 19 of its own in Iowa.
Context
Until 2022, outside Brazil and Chile there were no AWS, Google, Microsoft or Oracle regions in Latin America, so workloads from other countries were hosted abroad. Mexico's regions changed that for the region's second-largest market.
The choice of region determines latency, price, available services and compliance with data laws. It is worth reviewing service by service.
Figure 1 · Hourly price of the same instance by region
index, Virginia = 100Table 1 · Regions of the four major providers in Latin America
| Country | AWS | Google Cloud | Microsoft Azure | Oracle |
|---|---|---|---|---|
| Brazil | São Paulo (2011) | São Paulo (2017) | Brazil South (2014); Brazil Southeast (2020), restricted access | São Paulo (2019); Vinhedo (2021) |
| Mexico | Mexico Central (2025) | Querétaro (2024) | Mexico Central (2024) | Querétaro (2022); Monterrey (2023) |
| Chile | Announced for end of 2026 | Santiago (2021) | Chile Central (2025) | Santiago (2020); Valparaíso (2023) |
| Colombia | — | — | — | Bogotá (2023) |
| Venezuela and Panama | — | — | — | — |
Business implications
Check the region's catalogue before designing. GPUs, AI models and some managed services are not available in every region. Whatever is missing forces the use of another region, with effects on latency and data residency.
Include regional pricing in the business case. Across the instance types reviewed, São Paulo costs between 54% and 62% more than Virginia, and Mexico 5% more. It should be modelled as an explicit assumption of operating cost.
Document where data is replicated. Some regions copy information outside the country. In Azure, Brazil South is listed as replicated to the United States, while Chile Central and Mexico Central store data at rest in-country.
Methodology & data
Review of the official region, pricing and terms pages of AWS, Google Cloud, Microsoft Azure and Oracle, consulted on 28 September 2026. Prices are for on-demand instances with Linux and shared tenancy; the index and monthly equivalents are own calculations.
Catalogues, prices and terms change often; the data hold for the date of consultation.
References
Amazon Web Services. Regions and Availability Zones; Amazon EC2 On-Demand Pricing (consulted 28 September 2026). · Amazon (7 May 2025). Amazon to Invest More Than $4 Billion to Launch Infrastructure Region in Chile. · Google Cloud. Cloud locations; Compute Engine regions and zones; Generative AI locations (consulted 28 September 2026). · Microsoft. Azure regions list; Microsoft Datacenters (consulted 28 September 2026). · Oracle. Public Cloud Regions; Global Trade Compliance (consulted 28 September 2026).