Ascenty plans $6B AI data center complex in São Paulo state

This is one of the largest AI‑specific infrastructure announcements yet seen in Latin America. A R$30 billion (~$6 billion) complex dedicated to AI workloads in Sumaré and Vinhedo signals that hyperscale compute is no longer just a US, EU or East Asian story. ([piranot.com.br](https://www.piranot.com.br/2026/05/28/noticias/economia/ascenty-investe-us-12-bi-em-complexo-de-ia-em-sp/?utm_source=openai)) Ascenty, backed by Digital Realty and Brookfield, is effectively building a regional hub where global cloud and AI players can deploy supercomputers without sending all traffic back to Northern Virginia or Dublin.

For the AGI race, these regional AI “power plants” matter because they determine who can meaningfully participate in frontier‑scale training and inference. If Brazil can host 60–160 MW campuses optimized for liquid‑cooled AI superclusters, both domestic firms and multinational labs gain more geographical options for capacity expansion. That’s important for resilience, latency‑sensitive applications serving South American users, and potentially for regulatory arbitrage if different jurisdictions set different rules on model training or data localization.

It also highlights the emerging division of labor: hyperscalers like Microsoft and Oracle are expected anchor tenants, bringing model APIs and cloud platforms, while Ascenty specializes in power, cooling and physical infrastructure. Over time, such complexes could house not just general‑purpose clusters but specialized sovereign AI stacks, further fragmenting where and how frontier models are trained. The risk is that local permitting, grid constraints and environmental scrutiny could slow deployment just as AI demand spikes, testing Brazil’s regulatory agility.