The Global South faces a critical tension: artificial intelligence could help solve energy access and climate challenges, but the data centers powering AI are consuming electricity at unprecedented rates. As countries from India to Brazil to Kenya race to build AI infrastructure, they're confronting a fundamental question: can they harness AI's benefits without destabilizing their energy systems?

Why Is AI Infrastructure Exploding in Developing Regions?

Countries across the Global South are investing heavily in AI data centers to establish themselves as regional computing hubs and reduce dependence on US and Chinese infrastructure. The appeal is clear: AI could improve productivity in healthcare, agriculture, finance, and energy management while narrowing persistent gaps in development. Yet this expansion comes with a hidden cost measured in megawatts and gigawatts.

The scale of these projects is staggering. The United Arab Emirates is building Stargate UAE, a planned 1 gigawatt AI cluster in Abu Dhabi as part of a broader 5 gigawatt UAE-US AI Campus. Saudi Arabia's HUMAIN project, backed by the Kingdom's Public Investment Fund, plans to deploy up to 500 megawatts of AI infrastructure over the next five years with around $10 billion in investment. India is pursuing equally ambitious plans, with Yotta Data Services announcing a $2 billion AI supercluster and Google committing $15 billion to an AI data center hub in Visakhapatnam that will include a 1 gigawatt facility.

Southeast Asia, Africa, and South America are following suit. Malaysia's YTL-NVIDIA AI data center project received $4.3 billion in investment at a 600 megawatt facility. Indonesia is building its first renewable-powered AI data center park designed to scale to 500 megawatts. Kenya is developing a geothermal-powered data center planned to open at 100 megawatts with potential to reach 1 gigawatt. Brazil's Scala AI City is projected to expand from 54 megawatts to 1.8 gigawatts by 2033.

How Much Electricity Do These Data Centers Actually Need?

The energy demands are almost incomprehensible. The International Energy Agency estimates that data centers consumed around 1.5 percent of global electricity in 2024, or 415 terawatt-hours, and projects this could more than double to 945 terawatt-hours by 2030, roughly equivalent to Japan's entire annual electricity consumption. AI is the primary driver of this growth.

To put this in perspective, the planned 5 gigawatt G42 AI Campus in the UAE is estimated to consume 49.1 terawatt-hours annually, which exceeds the entire nation's annual electricity consumption. Even the Barakah Nuclear Energy Plant, which produces 25 percent of the UAE's total electricity, would be insufficient to power the new campus alone.

This mismatch between energy demand and available supply is forcing data center operators to consider captive or onsite power generation, raising concerns among experts that this could entrench fossil fuel consumption in regions already struggling with climate commitments.

What Are the Emerging Solutions to Power AI Data Centers?

The industry is experimenting with two distinct models to address the power constraint. The first involves massive long-term leases that bundle compute, power, and financing into integrated infrastructure platforms. The second pursues advanced nuclear reactors as a dedicated power source for AI facilities.

Hut 8's Beacon Point campus in Texas exemplifies the first approach. NVIDIA has reportedly emerged as the customer behind lease commitments worth as much as $50.2 billion at the one-gigawatt facility, according to reporting by the Financial Times, though NVIDIA neither confirmed nor denied the arrangement. The deal includes two 15-year leases with a combined base-term contract value of $19.6 billion, with the potential to reach $50.2 billion if all renewal options are exercised. The campus is being designed around NVIDIA's DSX reference architecture, which standardizes AI factory design and operations. Reporting has speculated that NVIDIA could sublease capacity to smaller cloud providers, allowing them to access large-scale capacity without independently signing multibillion-dollar leases.

The second model pairs advanced nuclear reactors directly with data center operations. Aalo Atomics and Crusoe announced a partnership intended to demonstrate a nuclear-powered AI data center by 2027 and begin deploying commercial 50-megawatt nuclear-powered AI factories by the end of 2029. The companies plan to deploy a Crusoe Spark modular data center running Crusoe Cloud at Idaho National Laboratory in 2027 as a proof-of-concept project. They then intend to deploy Aalo Pods, 50-megawatt-electric nuclear power plants, at Crusoe data centers by the end of 2029. Aalo's zero-power Critical Test Reactor reached criticality on July 4, 2026, sustaining a nuclear chain reaction without generating commercial electricity, marking a significant milestone toward commercial deployment.

Steps to Understanding Data Center Power Solutions

• Integrated Infrastructure Platforms: Companies are bundling compute architecture, power generation, financing, and operations into vertically coordinated systems rather than treating data centers as isolated buildings connected to whatever utility service is available.
• Long-Term Lease Structures: Anchor tenants with strong credit quality, like NVIDIA, can help make multibillion-dollar campuses financeable by committing to 15-year or longer leases that provide predictable revenue streams.
• Advanced Nuclear Deployment: Modular advanced reactors are being deployed directly at data center sites to provide dedicated, carbon-free baseload power without relying on grid infrastructure or fossil fuel generation.

What's the Catch for the Global South?

While these solutions show promise, they present distinct challenges for developing regions. The time mismatch is critical: new AI infrastructure is expected to come online within the next few years, but major energy infrastructure projects require longer planning, permitting, financing, and construction timelines. This means new AI data centers will likely draw power from each country's existing energy mix in the short term rather than from decarbonized future grids.

Data centers currently rely on a predominantly fossil fuel-based energy mix, and the International Energy Agency projects only a larger role for nuclear energy toward the end of the decade. For countries in the Global South pursuing industrial decarbonization while simultaneously seeking to reduce inequities in energy access, this timing creates a genuine dilemma. The value of AI for these countries will ultimately depend on whether it can be deployed without exacerbating existing constraints around energy access, grid reliability, climate vulnerability, and public investment.

The paradox is stark: AI could help optimize energy systems through grid management, renewable energy integration, predictive maintenance, and accurate forecasting of alternative energy sources. Yet its energy-intensive nature could simultaneously undermine the very energy ecosystems it's meant to improve, particularly in regions where energy access remains a fundamental development challenge.