The premise seemed obvious going in: somewhere in the South Pacific, the Atlantic, Africa, or Latin America, a small island with abundant green energy should be the natural next home for the world's data-center land rush. The research says otherwise. Almost every small island scanned for this letter is either too fossil-dependent, too small-gridded, or too disaster-exposed to host serious compute infrastructure today. The real opportunity is a landlocked country most data-center coverage still overlooks — producing roughly triple the electricity it consumes, entirely from hydro, with one firm alone floating $50 billion over 30 years to build there.
The premise going into this research was that a small island — cheap land, abundant sun or geothermal heat, eager for foreign investment — should be a natural home for the world's data-center land rush. The research doesn't support that premise, and the honest version of this letter is built around correcting it. Almost every small island scanned across the South Pacific, the Atlantic, and the Caribbean turns out to be too fossil-dependent, too small-gridded, or too disaster-exposed to host serious compute infrastructure today. The actual opportunity — real, dated, billions of dollars already committed — is a landlocked country most data-center coverage still overlooks entirely.
Start with where the money would seem most obvious to go, and isn't. Caribbean electricity currently averages around $0.25 per kWh — more than double the US rate, and above $0.40 in some island nations — because over 90% of Eastern Caribbean generation still runs on imported fossil fuel. The World Bank has committed roughly $500 million to Caribbean renewable transition, with $110 million specifically for Grenada, Saint Lucia, and Saint Vincent, but as of the most recent data only about 12% of the region's electricity comes from renewables. That is the opposite of a data-center cost advantage — it's currently one of the more expensive power environments in the hemisphere.
Pacific island states tell a similar story from a different angle. Fiji, Vanuatu, and the Solomon Islands have committed to 100% renewable power by 2030, and renewable capacity across the Pacific grew 30% between 2014 and 2022 — genuine progress, not greenwashing. But these remain small, isolated grids built for domestic consumption, not the hundreds of megawatts a single serious data-center campus requires, and cyclone exposure adds a real infrastructure-resilience risk any data-center operator's insurers would price in immediately.
Cape Verde is the most interesting near-miss in the Atlantic: renewables now cover roughly 25–33% of its electricity, up from a diesel-only baseline within the past decade, and household tariffs fell for a second consecutive year in 2026 as clean generation expanded. But the country's own 100%-renewable target date is 2040, not now, and there is no confirmed major data-center commitment in the country as of this letter — genuine progress, several years from being the story this letter is looking for.
Paraguay is landlocked, which immediately disqualifies it from anyone searching only for "islands" — but it is producing, right now, one of the largest untapped clean-energy surpluses on the planet. The country generates nearly 100% of its electricity from hydroelectric sources, primarily through two binational dams: Itaipú, a 14,000MW facility shared with Brazil whose annual output rivals China's Three Gorges Dam, and Yacyretá, shared with Argentina. Paraguay consumes only a fraction of its own share of Itaipú's output — its half could power the country roughly three times over — with the historical surplus sold cheaply to Brazil under a decades-old treaty that Paraguayan officials increasingly argue has undervalued the resource for fifty years.
That imbalance is now the basis of a real, dated, and large investment story. In March 2026, the foreign ministries of Paraguay and Brazil announced substantive progress renegotiating Annex C of the Itaipú Treaty — the pricing and distribution mechanism governing the dam's output. In the same window, US firm X8 Cloud floated plans to invest as much as $10–50 billion over 30 years in what it describes as South America's largest AI data-center complex. This isn't a single speculative announcement: Hive Digital Technologies is already building, not merely planning — its Yguazú facility completed Phase 1 in April 2025 and Phase 2 in September 2025, is expanding to 400MW with Phase 3 targeted for Q3 2026, and has publicly stated the expansion is shifting the site's focus from Bitcoin mining toward AI compute. In May 2026, Paraguayan President Santiago Peña signed a government-to-government accord with Taiwan specifically to pair Taiwanese technology with Paraguayan energy for a major AI center.
Every other candidate in this research had potential. Paraguay has the treaty renegotiation, the phased construction already two-thirds complete, the signed government accord, and investment-grade credit ratings from all three major agencies, simultaneously, in the same year.
Cheap hydro is the headline, but it isn't the whole case. Paraguay holds investment-grade sovereign credit ratings from all three major agencies — a genuine rarity among emerging-market energy exporters, and the kind of detail that determines whether a $50 billion, 30-year commitment is underwritable at all. Its territorial tax system means income is taxed only where it's earned within the country, a real structural advantage for a capital-intensive, internationally-financed data-center buildout. A simplified corporate formation framework (the EAS, Empresa por Acciones Simplificada) lowers the administrative barrier to entry. And its central South American location provides workable connection latency to markets across the continent — not the ultra-low latency of a coastal hub, but workable for the AI training and batch-processing workloads that dominate current data-center demand growth.
The honest constraint, and it's a real one: Paraguay's national grid can currently supply only about five gigawatts, a ceiling its own industry minister has acknowledged publicly. A single mega-project at the scale X8 Cloud has floated could consume a meaningful share of that capacity on its own. The regulatory framework for electrointensive industries is still being built in real time by the national electricity administrator (ANDE), not yet a mature, tested system. Critics have also noted, fairly, that early-stage investment figures in fast-moving frontier markets like this one often prove larger in press coverage than in what actually gets built.
Two other real, non-speculative cases are worth naming directly, because they show this isn't only a South American story. Kenya's Rift Valley geothermal resource has already attracted a genuine, large commitment: Microsoft and G42 committed $1 billion to a geothermal-powered data center in the country, leveraging a power source that — unlike solar or wind — runs continuously regardless of weather, a real technical advantage for the uninterrupted load a data center requires. Iceland remains the mature, proven benchmark this entire category measures itself against: a grid sourced entirely from geothermal and hydro, a cold climate enabling free air-cooling that cuts operating costs by up to 40%, and years of established hyperscaler presence from Google and Meta, the latter through a direct geothermal partnership with Fervo Energy. Iceland isn't a "new opportunity" anymore — it's the finished proof that this entire model works, which is precisely why Paraguay's current, earlier-stage version of the same story is the more interesting entry point today.
The broader pattern connects directly to why any of this research mattered in the first place. Traditional data-center hubs have hit real physical limits: interconnection queues in primary US markets now stretch beyond four years, Ireland imposed a moratorium after its data centers began consuming more power than all its urban homes combined, and Northern Virginia's grid has reached capacity. That structural wall is what's pushing serious capital toward "follow the electrons" strategies — going to wherever cheap, abundant, already-clean power genuinely exists, rather than trying to build new generation in already-saturated markets. Paraguay's surplus was created by a fifty-year-old treaty nobody originally built for this purpose; it is now one of the more consequential accidents of infrastructure timing in the current AI buildout.
The comprehensive answer to "which small islands have the greatest scope for green-energy data centers" turns out to be: almost none of them, not yet — the Caribbean is currently power-expensive, Pacific SIDS are grid-constrained and disaster-exposed, and Cape Verde is genuinely promising but years from ready. The real, current, capital-backed opportunity is Paraguay, a landlocked country outside the "island" framing entirely, with a real treaty renegotiation underway, a facility already two-thirds built, a signed government-to-government accord, and investment-grade credit — not a pitch deck. Kenya's geothermal story and Iceland's mature, proven model round out the genuine cases; everywhere else scanned in this research is future potential, not present opportunity. The risk to this thesis: Paraguay's own five-gigawatt grid ceiling is real and near-term binding — the difference between the $50 billion headline figure and what actually gets built will depend heavily on how fast grid capacity, the Itaipú treaty renegotiation, and ANDE's regulatory framework for electrointensive industries all resolve, none of which are fully settled as of this letter.