Rio AI City — Elea’s 3.2 GW Hyperscale Data Center Campus Transforming Latin America

Updated March 2026

When Mayor Eduardo Paes took the stage at Web Summit Rio in April 2025 and announced Rio AI City, he did not unveil a modest pilot project or a speculative feasibility study. He revealed a hyperscale digital campus developed by Elea Data Centers with a target capacity of 3.2 gigawatts at full build — enough electrical power to rank among the ten largest data center campuses on Earth. The project’s ambition is nothing less than transforming Rio de Janeiro into the largest data center hub in Latin America, positioning the city as a critical node in the global AI infrastructure supply chain alongside established hubs in Northern Virginia, Singapore, Amsterdam, and Frankfurt.

The Strategic Vision Behind Rio AI City

The announcement of Rio AI City arrives at a moment when demand for AI computing infrastructure is outstripping supply globally. The training and deployment of large language models, computer vision systems, and generative AI applications require data center capacity measured not in megawatts but in gigawatts — a scale that only a handful of locations worldwide can provide. Rio’s candidacy for this role rests on several structural advantages that distinguish it from competing locations in Latin America and beyond.

First, Rio is a primary landing point for multiple submarine cables connecting South America to Central and North America, Europe, and Africa. These cables provide the high-bandwidth, low-latency international connectivity that data center tenants require. A hyperscale facility sited near submarine cable landing stations eliminates the latency penalty associated with backhauling data to inland locations, a critical advantage for AI inference workloads where milliseconds of additional latency degrade user experience.

Second, Brazil’s National Data Center Policy, scheduled for launch in May 2025, promises tax incentives, legal security, and sector-specific rules designed to attract international data center investment. This policy framework, developed in parallel with the $4 billion National AI Plan of 2024, creates a regulatory environment tailored to the needs of hyperscale operators. Tax incentives for data center construction and equipment imports directly reduce the capital expenditure required for facilities at the scale of Rio AI City.

Third, Rio’s position as the headquarters city for Petrobras, Eletrobras, BNDES, and other major state enterprises creates institutional demand for data center services from organizations that prefer to keep their computing infrastructure within the metropolitan area for reasons of security, latency, and regulatory compliance. This captive demand provides a baseline utilization floor that reduces the financial risk of speculative construction.

Elea Data Centers: The Developer Behind the Campus

Elea Data Centers brings specialized expertise in designing and operating carrier-neutral data center facilities in Brazil. The company’s approach to Rio AI City follows a phased development model that manages capital risk while allowing capacity to scale in response to market demand. Phase RJO1 is already operational, providing the initial anchor capacity that establishes the campus as a functioning facility rather than a paper project. Phase RJO2, with 80 MW of capacity, is scheduled for delivery in 2026, representing the first major expansion block.

The phased approach is standard practice in hyperscale data center development, where the total capital investment for a 3.2 GW campus would run into the tens of billions of dollars if constructed simultaneously. By delivering capacity in increments, Elea can secure tenants for each phase before committing capital to the next, maintain construction quality by avoiding the coordination challenges of simultaneous multi-hundred-megawatt builds, and incorporate evolving cooling and power distribution technologies as they become commercially available.

The campus design must address Rio’s tropical climate, where average temperatures range from 22 to 30 degrees Celsius year-round with high humidity. Traditional data center cooling approaches designed for temperate climates are less efficient in tropical environments, and Elea’s engineering team is implementing advanced cooling architectures that may include direct liquid cooling for high-density AI computing racks, evaporative cooling systems optimized for humid conditions, and waste heat recovery systems that redirect thermal energy to useful purposes rather than simply rejecting it to the atmosphere.

Capacity in Context: What 3.2 GW Means

To appreciate the significance of 3.2 GW of data center capacity, consider that the average megawatt of data center capacity can support approximately 200 standard server racks or 50 to 100 high-density AI computing racks depending on the specific hardware deployed. At 3.2 GW, Rio AI City would theoretically support hundreds of thousands of server racks — a figure that would place it among the largest concentrations of computing power in the Southern Hemisphere.

For context, the COR Operations Center currently operates 84 servers with approximately 10 petabytes of storage to manage the city’s entire smart city sensor network. The computing capacity available at full Rio AI City buildout would exceed COR’s current installation by orders of magnitude, enabling AI workloads that are currently only feasible at a handful of hyperscale campuses in North America and Asia.

The transition from megawatts to gigawatts represents a qualitative shift in what becomes possible. At megawatt scale, a data center serves as a hosting facility for enterprise applications and cloud services. At gigawatt scale, a campus becomes an AI training facility capable of running the months-long distributed computing jobs required to train frontier AI models. This distinction matters because AI training workloads are the fastest-growing segment of data center demand globally, and locations that can offer gigawatt-scale capacity with reliable power and cooling are in extremely short supply.

Energy Supply and Sustainability

Powering a 3.2 GW data center campus is itself a major infrastructure challenge. For reference, 3.2 GW exceeds the total electricity consumption of many mid-sized countries and represents a significant fraction of Rio de Janeiro state’s total generating capacity. Meeting this demand will require dedicated power infrastructure — likely including direct connections to the national transmission grid, on-site substations, and potentially dedicated generating capacity.

Brazil’s energy matrix provides a structural advantage here. Approximately 83 percent of Brazil’s electricity comes from renewable sources, primarily hydroelectric power, with growing contributions from wind and solar. This means that a data center campus in Rio can credibly claim a substantially lower carbon footprint per megawatt than equivalent facilities in markets that rely heavily on fossil fuel generation. For AI companies facing increasing pressure from investors and customers to reduce the carbon intensity of their computing operations, this renewable energy advantage creates a competitive pull toward Brazilian data center locations.

The smart energy grid initiatives being deployed across Rio, including solar mandates for new construction and smart metering infrastructure, create an ecosystem-level context for energy management that benefits data center operations. Smart grid data about real-time generation capacity, transmission line utilization, and demand patterns can inform data center operators about optimal times to run power-intensive training workloads, potentially shifting flexible computing loads to periods of peak renewable generation and lowest grid carbon intensity.

Elea’s pursuit of sustainability credentials, mirroring the COR facility’s LEED certification pursuit, reflects industry trends where data center operators increasingly compete on environmental performance. Major cloud providers and AI companies now include sustainability requirements in their data center procurement processes, making green certification and renewable energy sourcing commercial necessities rather than optional corporate social responsibility gestures.

Impact on Rio’s Technology Ecosystem

The announcement of Rio AI City has ripple effects across the city’s technology ecosystem that extend far beyond the direct employment and investment associated with the campus itself. A hyperscale data center campus of this magnitude will attract cloud service providers, AI companies, managed service providers, and systems integrators that cluster near major data center installations to minimize latency and maximize access to computing resources.

This clustering effect is well-documented in established data center markets. Northern Virginia’s data center corridor, which hosts approximately 300 data centers with over 2.5 GW of capacity, has attracted thousands of technology companies that benefit from proximity to cloud infrastructure. Dublin, Amsterdam, and Singapore have experienced similar clustering effects, where data center density begets technology company density in a self-reinforcing cycle.

For Rio’s existing startup ecosystem, which ranks sixth in Latin America according to Startup Genome, the availability of world-class computing infrastructure within the metropolitan area removes a significant constraint. AI startups that previously had to rely on cloud regions in Sao Paulo, Virginia, or elsewhere can now access equivalent computing resources with lower latency and potentially lower costs. The Porto Maravalley innovation hub, already home to Google and Meta offices, gains additional appeal as an AI startup location when combined with local hyperscale computing access.

The tech workforce pipeline from UFRJ, PUC-Rio, and FGV will also benefit. Data center operations, AI model development, cloud engineering, and related disciplines will see increased demand, creating career pathways that keep technical talent in Rio rather than losing it to Sao Paulo, Lisbon, or North American tech hubs. Elea and its future tenants will need engineers, technicians, security specialists, and operations staff at scale, creating direct employment opportunities in the thousands.

Competitive Positioning in Latin America

Rio AI City enters a Latin American data center market that is growing rapidly but remains small by global standards. Sao Paulo currently dominates the Brazilian data center market, with Campinas emerging as a secondary hub due to its position on the fiber optic backbone connecting Sao Paulo to international submarine cable landing points. Santiago, Bogota, and Mexico City are establishing data center clusters of their own, competing for the same multinational cloud providers and AI companies.

Rio’s pitch to these potential tenants rests on three differentiators: submarine cable proximity (reducing latency to international networks compared to inland locations like Sao Paulo or Campinas), the municipal government’s active support through smart city infrastructure and favorable regulatory positioning, and the sheer scale of the 3.2 GW target capacity, which signals a commitment to the kind of long-term infrastructure build that hyperscale tenants require before committing to a location.

The competition with Sao Paulo’s tech ecosystem is not zero-sum. As total Latin American data center demand grows, both cities can expand simultaneously. However, Rio AI City’s success would shift the balance of the region’s technology infrastructure map, ending Sao Paulo’s near-monopoly on premium data center capacity in Brazil and creating redundancy that benefits the entire national technology ecosystem. For international companies considering Brazilian operations, having two world-class data center locations provides resilience against localized disruptions and optionality for workload placement.

The National Policy Tailwind

Rio AI City’s development coincides with an unprecedented level of federal policy support for data center and AI infrastructure in Brazil. The $4 billion National AI Plan of 2024, the EBIA (Brazilian AI Strategy) framework launched in 2021, and the forthcoming National Data Center Policy collectively create a policy environment designed to accelerate precisely the kind of investment that Rio AI City represents.

The National Data Center Policy’s tax incentives are particularly significant for a project of this scale. Data center construction involves massive capital expenditure on specialized equipment — power distribution systems, cooling infrastructure, server racks, fiber optic connectivity, and security systems — much of which is imported. Tax relief on these imports directly reduces the cost per megawatt of delivered capacity, improving the project’s financial viability and allowing Elea to offer more competitive pricing to tenants.

Legal security provisions in the national policy address a concern that has historically slowed data center investment in Brazil: regulatory uncertainty around data sovereignty, taxation of digital services, and municipal permitting for large-scale industrial facilities. By establishing clear, sector-specific rules at the federal level, the policy reduces the risk of post-investment regulatory changes that could alter the economic assumptions underlying a multi-decade infrastructure project.

What to Watch Next

The near-term milestone for Rio AI City is the delivery of Phase RJO2’s 80 MW capacity in 2026. The success of this phase — measured by construction timeline adherence, tenant uptake, and operational performance — will determine the pace and scale of subsequent expansion phases. If RJO2 delivers on schedule and attracts marquee tenants, the pathway to the full 3.2 GW buildout becomes more credible and more likely to attract the co-investment and long-term tenant commitments needed to sustain construction across what will likely be a decade-long development program.

The intersection of Rio AI City with the city’s broader smart city infrastructure creates a unique positioning opportunity. No other hyperscale data center campus in Latin America is being built in a city that simultaneously operates a 10,000-camera smart city network, an AI-powered traffic management system, a 5G pilot program, and an extensive IoT sensor network. This combination of AI infrastructure supply (computing capacity) and AI infrastructure demand (smart city applications) creates a living laboratory for urban AI that could attract research partnerships, government pilot programs, and technology companies seeking real-world testing environments for their products.

External Resources

• Global Data Center Hub — Rio AI City Analysis — Detailed coverage of the announcement and market context
• Elea Data Centers — Developer profile and campus specifications