Oracle, Nvidia, and Elea Join Forces to Build Latin America’s Largest AI Data Hub in Rio

How the Oracle-Nvidia-Elea Partnership Will Shape the Rio AI City Project

Rio AI City, an ambitious data center cluster led by Elea Data Centers, has gained fresh momentum with Oracle and Nvidia formalizing their involvement through a memorandum of understanding (MOU). This agreement was signed during Rio Innovation Week, in the presence of mayor Eduardo Paes and other local officials. Earlier reports indicated Nvidia was already in talks to join the initiative.

The MOU builds on a prior agreement involving Rio de Janeiro city, the federal government, and agencies such as Finep, BNDES, and Eletrobras. Rio AI City aims to reach approximately 1.5GW of capacity by 2028, scaling up to 3.2GW by 2032, with investments potentially reaching US$65 billion. The first phase is expected to be operational by 2026.

Intentions Behind the Agreement

The MOU is primarily a statement of intent rather than a firm investment commitment. No specific figures or hardware commitments have been made public yet, largely because tenant demand remains uncertain. The project is still in the customer prospecting phase, which will advance once key licensing and energy supply issues are resolved.

The success of Rio AI City also depends on broader factors, including whether major tech companies like Microsoft, AWS, Google, or other AI leaders decide to invest. Their decisions will be influenced by Brazil’s AI regulatory framework and potential tax incentives.

Oracle will supply servers powered by Nvidia’s AI GPUs and software, as Oracle is a client of Nvidia. Elea is managing the overall infrastructure—covering space, energy needs, and commercial sales. Mayor Paes strongly supports the project, recognizing its potential economic impact.

Márcio Aguiar, Nvidia’s head of enterprise for Latin America, explained that the city approached Nvidia to assist with energy assessments for GPUs. These studies will help determine how many GPUs the data center can support as it scales.

Market data suggests that 1,000 GPUs require roughly 0.8GW of power, so a 3GW facility could support around 3,500 GPUs. This informs infrastructure planning, including the need for additional buildings or halls.

Elea’s CEO Alessandro Lombardi emphasized the collaboration with Nvidia to adapt engineering and design to computing needs, ensuring readiness once demand is confirmed.

Oracle and Nvidia: A Coordinated Effort

Oracle and Nvidia have an established partnership on several large projects. They recently agreed to develop 4.5GW of new data center capacity in the US for OpenAI’s Stargate initiative. They’ve also worked together on startup acceleration and supercomputer clusters in Latin America.

Oracle, despite entering the public cloud market later than some competitors, has invested heavily to catch up. It operates its own Oracle Cloud Infrastructure (OCI) and collaborates with other hyperscalers in multi-cloud and hybrid environments. Many enterprises use Oracle databases, including on-premises setups.

Alexandre Maioral, president of Oracle Brazil, noted that Rio AI City aligns with Oracle’s goal to provide advanced AI technology and data center infrastructure in Brazil and Latin America. The focus is on delivering secure, scalable, and sustainable solutions.

Market sources indicate Oracle has access to the latest generation GPUs, sometimes even more so than AWS. This makes Oracle an important player for companies struggling to find GPU resources elsewhere.

Lombardi pointed out that Rio de Janeiro lacked an Oracle cloud presence until now. He sees Oracle’s cloud as an agnostic foundation critical for robust machine learning models. OpenAI, for example, runs on Oracle’s stack, highlighting the importance of Oracle’s infrastructure for the region.

Not Just Oracle: A Broader Ecosystem

While Oracle is currently the “first mover,” discussions with other cloud providers are ongoing. Nvidia’s Aguiar emphasized that the agreement with Oracle does not confer exclusivity or preference.

Lombardi indicated that the Rio AI City project is nearly sold out due to strong demand from tech companies.

However, he highlighted a major barrier: Brazil’s 52% import tariff on equipment. The lack of a national data center policy offering tax exemptions (ReData) makes large-scale investments challenging.

Addressing these policy and tariff issues will be critical to unlocking the full potential of Rio AI City and attracting further investment.

For IT professionals interested in AI infrastructure and cloud developments, Rio AI City represents a significant case study on the intersection of government policy, energy management, and technology partnerships in emerging markets.

Explore more about AI infrastructure and cloud technologies with practical courses at Complete AI Training.