What can Chilean companies move to AWS in Santiago today?
AWS has opened its Chile Region to selected customers. For a bank or retailer, the first migration depends on local service availability and the cost of running two environments.
A Chilean retailer invited into Amazon Web Services’ new Region has a choice about its first migration. It could move an order database that uses services available locally, or wait until it can also move the system that recommends products at checkout. The right sequence depends on whether checkout still needs to call the old environment after the database moves. One test order, traced from payment through fulfilment and recovery, is more informative than the Region’s opening date.
On 11 November, AWS opened its Chile Region to selected customers, including BancoEstado, Cencosud and Ripley. AWS expects general availability by the end of 2026. Compute, storage, networking and databases form the initial offer. AWS puts Bedrock and SageMaker AI within the first year of launch. A company budgeting for local AI inference therefore needs to check its particular model, service and quota before committing to a cutover date. AWS launch announcement, 11 November 2026
Two suppliers can now compete for the same workload
Santiago already had a local cloud alternative. Microsoft opened its Chile Central region in June 2025 across three physically independent locations in the metropolitan area. AWS also had a presence before this launch. In 2023 it opened a Direct Connect location at SONDA’s Quilicura Q2 data centre, where companies could establish private connections to AWS. What changed this week is that early-access customers can run supported AWS services inside Chile. A company with an existing private circuit can now ask its carrier for a route and price to the new Region. Microsoft’s Chile Central announcement · AWS’s Quilicura announcement
For a retailer already operating on AWS abroad, keeping its database engine and deployment tools could reduce the work of moving. For a business already running in Azure in Chile, staying put may be cheaper even if a competing compute quote looks attractive. Both suppliers can be asked to price the same transaction volume, backup retention and recovery objective. A restore test will show whether the cheaper design actually meets the recovery time the business has promised.
How much does an extra month of migration cost?
Consider a buyer whose existing environment costs US$900 a day to keep running after the new one starts. Assume that cost remains fixed until it can retire the old environment. The table shows the additional cost of keeping it during the overlap. The new environment is common to all three cases. Migration labour and data-transfer charges would be priced separately.
| Time with both environments running | Additional cost of keeping the old environment |
|---|---|
| 2 weeks (14 days) | US$12,600 |
| 4 weeks (28 days) | US$25,200 |
| 8 weeks (56 days) | US$50,400 |
Extending the overlap from four to eight weeks adds US$25,200 under these assumptions. The buyer can compare that amount with the cost of separating one dependency earlier. The calculation changes if the old contract is prepaid and cannot be terminated. In that case an earlier cutover may release capacity for another workload without reducing the next invoice. The renewal date and the right to scale down matter as much as the quoted daily rate.
Which opportunities reach Chilean data centre operators?
Amazon describes its planned investment of more than US$4 billion over 15 years as covering construction, connection, operation and maintenance. That figure alone cannot tell a developer how many racks a customer will lease this quarter. A nearer-term opportunity for a Santiago operator is a customer retaining hardware for a payment or inventory system while it moves other services. That operator can quote a cabinet, two network routes and a staged exit date against a real requirement. AWS’s Chile investment plan
PowerlandMap’s public Chile inventory on 13 November contained 12 published supply records after excluding demonstrations. Seven were marked operational, two under construction and three in development. These counts describe records, which can represent buildings or phases. They give a buyer seven operating records to investigate. Only a dated operator quote can establish available space, power and handover for its equipment.
The next useful milestone for a wider group of buyers is general availability. For companies already admitted, the decision can come sooner. Before moving a live application, obtain its local service list and quotas, run a transaction and a restore from the intended locations, and agree the date on which the old environment stops costing money. If a critical service still runs abroad, price that route and test the extra trip while there is still time to change the design.
Cover illustration of Santiago and the Andes, created for this analysis.
The data behind the analysis
Follow Market Intelligence with sourced, dated evidence
Each analysis draws on qualified supply records, anonymised demand signals and evidence-based matching. The underlying records retain their sources, confidence levels and verification dates.
Related analysis
Jakarta’s JK1 expansion makes a case for local AI inference
Equinix is adding about 1,125 cabinets at JK1. For AI buyers in Indonesia, private cloud connections and local network access could matter as much as the space itself.
Norway’s data-centre bottleneck is becoming an allocation regime
Norway’s planned licensing regime and Statnett’s reservation pause turn grid access into a selection problem. The market must separate requested, reserved, connected and IT capacity.
Japan’s Data Centre Market After STTGDC’s ¥93bn Loan
STTGDC’s latest Tokyo financing highlights how Japan’s data centre competition is splitting between established customer locations and projects built around access to power.