AWS grows Trainium and Chile footprint

Amazon is pushing on two fronts at once. One is silicon — its Trainium AI chips. The other is geography — more physical cloud capacity in places like Chile. The news this week is that those two threads look more real, not less: Amazon’s Chile data-center project cleared a major legal obstacle, while AWS keeps making Trainium central to its AI story. (wncy.com) ### What happened in Chile? A court in Chile rejected an environmental challenge from residents and environmental groups who were trying to stop Amazon’s planned data-center complex in Huechuraba, on the northern edge of Santiago. That does not mean every controversy is over, but it does mean the project can k(wncy.com)e power line tied to the buildout. (wncy.com) ### What is AWS actually building there? AWS announced in May 2025 that it plans to launch a new South America (Chile) Region by the end of 2026. That region is supposed to open with three Availability Zones — basically three separate data-center clusters close enough to work together but far enough apart for (wncy.com) maintenance. (aws.amazon.com) ### Why does a Chile region matter? For customers in Chile and nearby markets, local infrastructure means lower latency and easier data residency compliance. That matters for banks, governments, retailers, and anyone building AI applications that need fast response times and want data to stay in-country or at least in-region. AWS has framed the Chile region as its third in Latin America, after São Paulo and Mexico. (aws.amazon.com) ### Where does Trainium fit into this? Trainium is Amazon’s custom AI accelerator — the chip family AWS wants customers to use for training and, increasingly, deploying large models. The basic pitch is simple: if Amazon can supply the chips, the servers, the clusters, and the cloud platform, it does not have to rely as heavily on Nvidia f(aws.amazon.com)argins. (aboutamazon.com) ### Is Trainium still a side project? Not really. Amazon said its chips business — Graviton, Trainium, and Nitro together — is now running above a $20 billion annual revenue rate, with triple-digit year-over-year growth. AWS also said Bedrock processed more tokens in Q1 2026 than in all prior periods combined, which tells you AI demand is no longer a future promise inside AWS. It is already showing up in usage and revenue. (finance.yahoo.com) ### What is the clearest proof customers will use it? Anthropic is the biggest proof point. In April 2026, Anthropic said it had expanded its deal with Amazon to secure up to 5 gigawatts of compute for training and deploying Claude, including new Trainium2 capacity coming online in the first half of 2026 and nearly 1 gigawatt of Trainium2 and Trainium3 capacity by the end of 2026. That is not a pilot. That is industrial scale. (anthropic.com) ### So why connect Chile and Trainium? Because AWS’s AI strategy is no longer just “rent GPUs in Virginia.” It is becoming a map-and-silicon strategy — put more regional capacity closer to customers, and fill the biggest AI clusters with Amazon-designed chips. Chile helps with reach and locality. Trainium helps with economics and supply control. The catch is t(anthropic.com)ave to prove they can keep winning real workloads. (wncy.com) ### Bottom line? AWS is trying to own more of the AI stack, from chip to region. The Chile ruling helps the physical footprint grow. Trainium is the part that could make that footprint much more profitable. (wncy.com)