The Topic

I Squared Capital Buys Into Brazil's AI Data Center Boom Why a $55 billion infrastructure manager is betting on Elea, energy, and one of Latin America's biggest digital markets

Abstract

The conventional read on data center deals is that they are about real estate and chips: secure land, fill it with servers, win cloud customers. The more revealing story in I Squared Capital's agreement to acquire a majority stake in Brazil's Elea Data Centers is that the binding constraint has quietly shifted to electricity. Elea's pitch, and the reason a global infrastructure investor is writing a large check, rests on access to gigawatt-scale renewable power in a country where it is said to be abundant. The mechanism is straightforward: secure "powered land" and renewable supply first, then build high-density capacity for artificial intelligence workloads on top. The implications are a multi-billion-dollar build-out, a test of whether energy-rich emerging markets can host frontier AI, and a reminder that many headline figures here are pipeline, not installed capacity.

Keywords: data centers; artificial intelligence; renewable energy; Brazil; infrastructure investing; I Squared Capital; Elea; powered land

1. Why This Matters Now

If you follow technology, you have heard that the artificial intelligence boom is running into a wall, and that the wall is increasingly power, not chips. This deal is a clean illustration. On April 28, 2026, I Squared Capital, a global infrastructure manager with roughly $55 billion in assets, agreed to acquire a majority stake in Elea Data Centers, with announcements issued from Miami and Rio de Janeiro. Elea is a carrier-neutral platform in Brazil, meaning it stays independent of any single telecom carrier and lets many networks interconnect inside its facilities. The right way to think about this is not as another property deal, but as a bet that whoever controls cheap, clean electricity controls the next phase of AI infrastructure.

2. Why This Matters for Tomorrow

Over the next few years, the leverage in this industry is migrating. It is moving away from who can buy the most advanced chips and toward who can secure energy, grid connections, and land at scale. That shift reshapes who wins. Infrastructure funds with patient capital and project-finance expertise become natural owners of AI's physical layer, while the moat increasingly lies in long-dated power access rather than software. Geography matters too. I Squared has invested in Latin America since 2015, deploying more than $3 billion across seven regional investments, and this deal extends that franchise into the most capital-intensive corner of digital infrastructure. If energy access is the real bottleneck, then countries with abundant renewable generation become strategically important in a way they were not when data centers clustered near a handful of established hubs. Expect competition, and eventually regulatory attention, to follow the power.

3. The Big Idea in Plain English

Think of an AI data center less like an office building and more like an aluminum smelter: an enormous, always-on consumer of electricity where the economics live or die on the power bill. Elea's strategy is to lock down the electricity first, then build the computing on top of it. In the old world, you picked a city near customers and fiber, then arranged power as a detail. In the new world Elea is betting on, you start from where gigawatts of renewable energy are available and build the digital infrastructure to meet it. The product being sold is, fundamentally, energy converted into computation.

4. How It Works (At a High Level)

From a customer's perspective, the flow is simple: an enterprise or a hyperscale company (a giant cloud operator like the ones running global AI services) needs computing capacity, signs up for space and power inside a facility, and plugs into the networks already present there. Behind that simplicity sits a stack of components.

1. Powered land. Elea reports over 300 MW of "powered land," its term for sites with planned grid access, plus over 1 GW of capacity currently in development. Delivering that pipeline is expected to require more than US$10 billion of capital, which is why a deep-pocketed infrastructure investor is involved.

2. Carrier-neutral campuses. Elea operates a nationwide network of interconnected campuses in key Brazilian metros including São Paulo, Rio de Janeiro, and Brasília, serving enterprise and hyperscale customers. Sources report between seven and nine interconnected sites, with the discrepancy likely reflecting the pace of recent expansion. Because the platform is carrier-neutral, many networks meet inside, which improves connectivity and customer choice.

3. High-density build-out. The capital from I Squared is intended to fund expansion and accelerate large-scale, high-density AI infrastructure across Brazil. High-density means racks packed with power-hungry AI chips, which demand far more electricity and cooling than traditional servers.

4. A flagship project. Elea is developing "Rio AI City," a planned multi-gigawatt AI campus in Rio de Janeiro intended to run on renewable energy. The headline capacity figure is itself contested: the deal announcement cites a flat 3 GW, one independent write-up reports up to 3.2 GW, and Elea's own materials reference 1.5 GW scalable to 3.2 GW. Whether that capacity is included within the "over 1 GW in development" figure or is separate and additive is not clarified in the deal materials, which leaves total planned capacity indeterminate.

5. Governance continuity. The deal partners with Elea's founder-led team, including President and Founder Alessandro Lombardi and I Squared Chief Investment Officer Gautam Bhandari, alongside continuing shareholder Piemonte Holding.

5. What Changes Because of This

For companies, this lowers a real barrier. Building gigawatt-scale capacity is brutally capital-intensive, and pairing a founder-led developer with an infrastructure manager that can underwrite billions changes what is buildable. Hyperscale customers that want AI capacity in Latin America gain a larger, better-funded supplier, while smaller regional developers without that backing are relatively disadvantaged.

For work and workflows, the centrality of energy reshapes who matters inside these projects. Power-purchase negotiators, grid-interconnection specialists, and high-density cooling engineers move from supporting roles to the critical path, ahead of the people simply leasing floor space.

The concrete, near-term example is Rio AI City itself, a stated plan rather than operating capacity, but one that signals the scale of ambition and is already drawing the deal's headlines. The medium-term, directional read: if this thesis plays out, expect more AI data center capacity to migrate toward energy-rich emerging markets over the next several years, rather than concentrating only in the United States and a few European hubs. It is worth noting that the materials do not state whether the development pipeline is confined to Brazil or includes planned capacity elsewhere in the region, even as the deal is framed as a broader Latin America franchise expansion. For end users, this is mostly invisible plumbing, but it determines whether AI services can scale affordably across the region.

6. Tensions, Risks, and Open Questions

Pipeline vs. installed. Most of the eye-catching numbers describe what could be built, not what exists. "Powered land" is undefined, so over 300 MW may mean land with planned grid access rather than energized, contracted capacity. Treat these as future-oriented figures.

Energy as savior vs. energy as risk. The founder's thesis is that AI is constrained more by access to energy than by technology, and Brazil offers gigawatt-scale renewable power. That framing is a stated opinion, not a neutral conclusion. Independent observers broadly agree energy access is a genuine constraint, but they also flag grid, transmission, and permitting risks, plus competing gigawatt-scale projects in Brazil.

The renewable claim, unverified. I Squared and Elea describe Rio AI City as designed to run on renewable energy, and the thesis leans heavily on that. But no independent audit, power-purchase agreement, or generation detail has been disclosed, so the claim survives as the company's stated design intent, not as a verified operational fact.

Announced vs. closed. The deal is agreed, not done, and remains subject to customary closing conditions. Financial terms, including valuation and the exact majority percentage, were not disclosed, so outsiders cannot yet judge the price or precise control structure.

7. Conversation Hooks

• "The interesting thing about this deal is the bottleneck moved from chips to electricity, so an infrastructure fund, not a tech company, is buying in."
• "I Squared has been in Latin America since 2015, so this isn't a tourist investment. They're building a franchise, not making a one-off bet."
• "Watch the language carefully: a lot of these numbers are 'powered land' and pipeline, not capacity that's actually running."
• "The renewable-energy claim is the load-bearing wall of the whole thesis, and it hasn't been independently verified yet."

8. If You Remember Three Things…

• I Squared agreed to buy a majority of Brazil's Elea to build AI data centers, betting that energy access, not technology, is the real constraint.
• The headline numbers (300+ MW of "powered land," 1+ GW in development, $10 billion-plus to deliver) are mostly future pipeline, not installed capacity.
• Watch whether "Rio AI City" and the renewable-power promise actually materialize, and on what schedule, since the deal hasn't even closed yet.

9. For the Nerds

For the nerds

The phrase doing the most quiet work here is "powered land," which sits ambiguously between a land bank with planned grid access and genuinely energized, contracted capacity. That ambiguity matters because high-density AI build-outs are gated by interconnection queues and firm power delivery, not by square meters, and Brazil's transmission and permitting realities are non-trivial constraints on converting planned megawatts into deliverable ones. A second open question is whether the over 1 GW "in development" and the multi-gigawatt Rio AI City overlap, double-count, or sum, which leaves total planned capacity genuinely indeterminate. There is also a financing structure worth watching: pairing project-finance-style infrastructure capital with founder-led development is increasingly the template for AI's physical layer, but it concentrates execution risk in the power and interconnection workstreams. The frontier question is whether energy-rich emerging markets can clear those grid and supply hurdles fast enough to host frontier-scale training and inference, rather than only secondary workloads.