$600 Million, 450 Megawatts, One Tier IV Facility: Inside the Caribbean's First Real AI Infrastructure Race

From Declarations to Poured Concrete

The last eighteen months of Caribbean AI news read like a stack of signing ceremonies. A sovereignty pledge in Bangkok. An ethics roadmap in Santo Domingo. Regional associations, forums, and buildathons. All of that mattered, but none of it moved a single server into a rack. The story that matters to an investor reading this in July 2026 is different: two Caribbean governments have now committed to physical AI infrastructure with named partners, dated timelines, and megawatt figures attached.

On 22 October 2025, at the National Palace in Santo Domingo, the Dominican Republic signed a partnership with NVIDIA to build a Center of Excellence in Artificial Intelligence, the first such hub in Central America and the Caribbean. The agreement commits to training more than 1,000 Dominican professionals by the end of 2026 through an AI Academy run jointly with NVIDIA's Deep Learning Institute, and to a National AI Factory intended to give the country its own sovereign compute base rather than renting capacity abroad. NVIDIA's Calista Redmond called it a partnership focused on education and industrial innovation. President Luis Abinader framed the goal more bluntly: positioning the country as a technology creator rather than a consumer.

Four months later, on 23 February 2026, Abinader signed Decree 113-26 designating a $500 million Google digital exchange hub a matter of national priority. The facility, a 7,000 square metre site, will host new submarine cable landing points connecting the Dominican Republic directly to South Carolina and Virginia, tripling the country's direct cable links to the United States and increasing its fiber pair count roughly tenfold. Google's Brian Quigley described the goal as making the Dominican Republic a strategic exchange point between North, Central, and South America. Construction begins in March 2026, with completion targeted for early 2027. Combined with the NVIDIA agreement, the Dominican Republic's committed AI infrastructure package now exceeds $600 million, a figure with no real precedent anywhere else in the Caribbean.

Trinidad's Bet, Signed Two Days Ago

Then, on 11 July 2026, Trinidad and Tobago signed its own set of agreements: a memorandum of understanding with Ernst & Young for a 300-megawatt data center, and a separate MOU with Florida-based Hummingbird AI Holdings for 150 megawatts of AI infrastructure. A third agreement involved Pinnacle Steel and Vanadium Corp. Combined, the two data center MOUs commit the country to 450 megawatts of new AI compute capacity, and the government projects the initiatives will generate more than 5,000 jobs.

To put that figure in context: 450 megawatts is a substantial new industrial load on a national grid the size of Trinidad and Tobago's, arriving at the same time environmental groups are raising concerns about the water these facilities will need for cooling in a country that already experiences chronic shortages and intermittent supply in parts of its network. Our sibling publication AI Trinidad & Tobago has covered the water and power tradeoffs of these specific deals in detail, and it is worth reading alongside this piece if Trinidad is the market you are underwriting.

What matters for the wider thesis is the sequencing. The Dominican Republic moved first, with a sovereign compute strategy backed by the world's dominant AI chipmaker and the world's largest cloud infrastructure company. Trinidad and Tobago moved second, several months later, with a narrower but still substantial data center bet built around US engineering and AI infrastructure firms rather than the hyperscalers themselves. Two different entry strategies into the same emerging category, from two different Caribbean governments, within the same twelve-month window, mark the opening moves of a regional competition for a genuinely scarce resource: hyperscaler and enterprise AI capital that, until now, has almost entirely bypassed the Caribbean.

The One Tier IV Problem

Here is the constraint that both deals now have to answer to. The entire Caribbean has exactly one Tier IV data center: Blue NAP Americas, in Curacao, a facility the Curacao government now owns outright after taking over the site. Every other operational facility in the region sits at Tier II or Tier III, a level the Caribbean Data Center Association and regional operators generally treat as the realistic near-term baseline, since Tier IV's full redundancy requirements are expensive to justify outside a handful of connectivity hubs. Cloud Carib, a managed service provider working with Blue NAP, has been building a federated cloud on the Gaia-X framework, with the first equipped racks going live around March and at least two data spaces (one government, one commercial) targeted for Curacao this year.

The reason this matters more than the headline megawatt figures is power. GPU-dense AI infrastructure needs electricity measured in gigawatts, not megawatts, and most Caribbean grids were built for tourism load and residential demand, not industrial-scale compute. Trinidad and Tobago's new 450 megawatts of committed capacity is a meaningful fraction of the national grid on its own. The Dominican Republic's National AI Factory carries the same underlying constraint, even with NVIDIA and Google's balance sheets behind it. None of the announcements to date change the physics of Caribbean electricity generation. They simply commit governments and companies to solving it on a deadline.

Guyana's economy, boosted by oil production, is projected to grow 16.3% in 2026 according to the World Bank's Caribbean Economic Outlook, while the rest of the region, per the Caribbean Development Bank, is expected to grow just 1.1% excluding Guyana. That gap is not incidental to this story. Oil-fueled Guyana and, to a lesser extent, gas-rich Trinidad and Tobago are the two CARICOM economies with genuine near-term capacity to add gigawatts of new generation. Everywhere else in the Caribbean, a serious AI data center bet has to solve the power question before it solves the compute question, and that is precisely the gap that has kept most of the region out of this race entirely.

Why the Money Skipped Everyone Else

Latin America and the Caribbean together receive approximately 1.12% of global AI investment, according to the Portulans Institute's October 2025 figures, against a combined regional share of roughly 6.6% of global GDP. That mismatch is well documented on this site already. What is less discussed is how concentrated even that thin slice has become. Lily Dash, founder of Future Caribbean, has made a comparable point about Africa: 83% of the continent's first-half 2025 AI startup funding went to just four countries, Nigeria, Kenya, South Africa, and Egypt, leaving the rest of a 54-country continent competing for scraps. The Caribbean pattern in 2026 looks similar in miniature. Two governments, the Dominican Republic and Trinidad and Tobago, are absorbing nearly all of the announced infrastructure capital, while Jamaica, Barbados, the Eastern Caribbean states, and most of the rest of CARICOM remain, for now, spectators to their own regional story.

That concentration is not necessarily a market failure. Hyperscalers and infrastructure funds go where the grid, the geography, and the government incentives line up, and the Dominican Republic's position as a subsea cable crossroads between three continents is a genuine structural advantage that smaller islands cannot replicate. But for an investor, concentration risk cuts both ways. Being early in the Dominican Republic or Trinidad and Tobago means being early in the two Caribbean markets actually building the physical layer. It also means a portfolio overweight in exactly two national grids, two sets of local politics, and two hurricane tracks.

The Layer Nobody's Pricing Correctly

Direct equity in NVIDIA, Google, Ernst & Young, or Hummingbird AI Holdings is not a Caribbean-specific trade; it is a global infrastructure trade that happens to have a Caribbean address. The more interesting entry points for regionally focused capital sit one layer down.

Power generation and grid upgrade financing serving data center clusters is the most direct exposure to the constraint that will determine whether these projects deliver on schedule. Water-efficient and closed-loop cooling technology, suited to island utilities rather than continental ones, addresses the specific environmental objection already surfacing in Trinidad. Colocation and edge computing infrastructure for markets too small to attract a hyperscaler directly, Jamaica, Barbados, the Eastern Caribbean, gives those territories a way to participate in AI infrastructure without needing to win a national-scale bid against the Dominican Republic. Submarine cable capacity resale, once the new Google-backed landing points are operational, is a distinct commercial opportunity from the construction spend itself. And co-investment alongside the Inter-American Development Bank and the World Bank, both of which have funded Caribbean data infrastructure before, offers a lower-risk structure than a solo bet on any single facility.

Smaller, faster capital shows a second path worth watching. In May 2026, Sagicor Financial Corporation committed $5 million to the University of the West Indies' new I-INSIGHT institute, funding a Sagicor UWI AI and Financial Services Hub set to roll out across all UWI campuses by August 2026. Five million dollars will not build a data center. It will, however, build the talent pipeline that eventually staffs one, and it demonstrates that meaningful Caribbean AI capital does not have to arrive at gigawatt scale to matter. The IDB's own AI Capital Exchange, launched through Invest Caribbean at Sustainability Week in Bridgetown, Barbados, the first time that forum has been held in the region, is built on the same premise: a debt capital pre-qualification platform sized for Caribbean deal flow, not for hyperscaler-scale transactions.

Risk Factors Investors Should Price Explicitly

Risk Factor	Severity	What It Means Here
Power grid capacity	High	Most Caribbean grids were not built for gigawatt-scale industrial load; generation and transmission upgrades are the real bottleneck behind every announced megawatt figure
Water availability for cooling	High	Directly at issue in Trinidad and Tobago, where chronic shortages predate the new data center commitments; other Caribbean islands face similar constraints
MOU-to-operation execution gap	Medium	A signed memorandum is not a running facility; treat announced timelines (March 2026, early 2027) as the earliest plausible case, not the base case
Capital concentration in two markets	Medium	The Dominican Republic and Trinidad and Tobago are absorbing nearly all announced infrastructure capital; a regional portfolio needs deliberate diversification beyond these two
Hurricane and climate exposure	Medium	Physical infrastructure investment in a hurricane corridor requires resilience design and insurance structuring from day one, not as an afterthought
Political and regulatory continuity	Low-Med	Decrees and MOUs can, in principle, be revisited by a future administration; multi-year infrastructure spend benefits from structures that survive an election cycle

The Timing Argument

This is the earliest point at which the Caribbean AI infrastructure story has been backed by real dated commitments rather than communiques. The Dominican Republic's package is nine months old. Trinidad and Tobago's is two days old as of this writing. Neither has yet produced a fully operational, revenue-generating facility. That is precisely the window in which the capital structure of these projects, who finances the power upgrade, who owns the cooling contract, who resells the new cable capacity, gets decided. Investors who wait for a ribbon-cutting will be negotiating against parties who already have the operating relationships in place.

StarApple AI, founded by Adrian Dunkley in 2023 as the first AI company established in the Caribbean, and recognised across the region as the work of the Caribbean's leading AI entrepreneur, built Maestro AI Labs specifically to track signals like these: the gap between a government decree and an operating data center, and the smaller, less visible capital that fills it. This analysis is supported by that ongoing tracking work, not by the press releases alone.

Related Reading Across the Caribbean AI Network

This analysis draws on and connects to ongoing coverage across the wider Caribbean AI network. Worth reading alongside it: