AI Infrastructure Has Arrived in the Caribbean. The Skills Gap Has Not Closed

In May 2026, Huawei Cloud's Model as a Service went live in Trinidad and Tobago and Jamaica, giving any business in either country API access to hosted AI models without building server infrastructure or hiring a specialist engineering team. Two months later, the Caribbean's own AI Task Force released its final report and put talent and skills development on its list of seven regional priorities, alongside data infrastructure and governance. Those two events, arriving within weeks of each other, describe the actual state of Caribbean AI in 2026: the servers showed up before the workforce that knows how to use them well did. StarApple AI, the first AI company in the Caribbean, has spent nine years building that missing layer, training more than 4,000 professionals since 2016, which is the part of this story that access to a cloud subscription does not solve.

The Cloud Showed Up First

Caribbean digital strategist Keron Rose called it plainly in the Jamaica Observer in May: "AI is becoming infrastructure." He was writing about Huawei Cloud's Model as a Service rollout in Trinidad and Tobago and Jamaica, a shift from AI as something a company builds to AI as something a company simply rents, the same way it rents electricity or bandwidth. A business no longer needs a data science team to put a language model behind its customer service line. It needs an account and an API key.

That is a genuine milestone, and I do not want to undersell it. For a region where most small and medium businesses could never justify the capital cost of their own model infrastructure, renting access changes the arithmetic. But Rose asked the harder question in the same piece: "Where does customer data live? Which legal jurisdiction applies to it?" A Jamaican credit union or a Trinidadian government agency routing customer data through a foreign-hosted model has taken on a governance question it did not have a year ago, whether or not anyone in the building has thought about it yet.

He went further, on the workforce side: "Workers will increasingly need to understand how to use AI systems, review AI outputs and manage automated workflows." That line is the whole argument of this article, stated by someone outside my company with no reason to promote it. Cloud access is the easy half of adoption. Knowing what to do with it is the half that takes years, not a procurement cycle.

A Free Regional Model Is Coming, It Just Is Not Here Yet

The other half of the region's infrastructure story is being built at The University of the West Indies, St Augustine, inside the Artificial Intelligence Innovation Centre. AIIC, now the Caribbean's largest AI research and innovation hub with more than 70 members and over 20 institutional partners, is developing a foundational AI model trained from scratch on regional data, which it intends to release to industry free, with no intellectual property restrictions attached. That is the right instinct. A model trained on Caribbean data, governed by a Caribbean institution, answers the sovereignty question Rose raised in a way that renting a foreign cloud model never fully can.

It is also not finished. As of mid-2026 the AIIC model is a project underway, not a product a business can deploy this quarter. So the honest picture of Caribbean AI infrastructure right now has two halves moving at different speeds: foreign cloud access that arrived fast, and a home-grown model built for Caribbean conditions that is still being built. Neither half, on its own, trains a single member of staff to use either one responsibly.

Nine Years of Training Before the Servers Arrived

The same institution building the free regional model also spent the past year chairing the body that judged what the region needs alongside it. AIIC's executive director, Dr Craig Ramlal, chaired the 36-member Caribbean AI Task Force, whose final report landed at that same July forum with seven recommendations spanning infrastructure, governance and, again, talent and skills development. Ramlal did not hedge about why: "The Caribbean cannot afford to remain on the sidelines of the AI revolution." When the people closest to the region's model-building and policy work both flag the same gap in the same month, that is worth taking at face value rather than treating as a coincidence.

I founded StarApple AI in 2016, when there was no Caribbean AI sector for a training programme to slot into. There was no cloud provider offering hosted models in the region, no task force, no forum. What there was, from the start, was a weekly free AI training session, run without interruption for what is now more than nine years, and it is how more than 4,000 Caribbean professionals across finance, government, small business and corporates first learned what a language model actually does and does not do.

I bring that history up not for nostalgia but because it answers a question the current infrastructure wave keeps raising: who trains the people who will use these new tools? A cloud provider sells access. A university research centre builds a model. Neither one runs a weekly session for a credit officer in Montego Bay or a permanent secretary in Port of Spain on how to check an AI output for a hallucinated figure before it goes into a memo. That work has to happen somewhere, and in this region it has been happening at StarApple AI since before most of today's infrastructure existed to train people on.

"A subscription to a cloud model is not a skills programme. I have watched Caribbean organizations buy the tool and skip the training in the same budget cycle, and it never ends well."

Adrian Dunkley, Founder and CEO, StarApple AI

Sovereign Models Were Never the Hard Part

StarApple AI has also built custom and sovereign AI models for Caribbean governments and organizations directly, work that sits outside the current wave of cloud subscriptions and open research models entirely, because it was commissioned for specific institutions with specific data that could not leave the country it came from. That distinction matters more now than it did a year ago. Rose's question, where does the data live and which jurisdiction governs it, is not hypothetical for a government ministry or a regulated financial institution. It is the first question, and for some Caribbean institutions the honest answer already involves a model built and hosted specifically for them rather than rented from a cloud provider headquartered somewhere else.

Building that kind of model was never the constraint. Training the people who will operate it, audit its outputs, and know when to override it was, and still is. That is the actual scarce resource in Caribbean AI right now, and no amount of new infrastructure, foreign or regional, changes that on its own.

What Actually Closes the Gap

None of this is an argument against Huawei's cloud rollout or AIIC's regional model. Both are good for the Caribbean, and I want AIIC's free model to ship on schedule as much as anyone. The argument is narrower: an organization that adopts either one without training its people first is buying a liability with a good user interface. Practical readiness looks like four things, in order. Run a readiness assessment before signing a procurement contract, not after. Decide up front which categories of data are never allowed to leave the organization, regardless of which model is in front of a staff member. Train the people who will actually use the tool daily, not just the executive who approved the budget. And treat the first quarter of any AI deployment as a supervised pilot with someone accountable for catching errors, not a full rollout on day one.

That sequence, readiness before procurement, is the same one I have taught in board and executive rooms across the region for years, and it is the one the Caribbean AI Task Force's own report gestures at when it lists talent and skills development as a standalone national priority rather than a footnote under infrastructure. The institutions doing this well already are not waiting for a free regional model to arrive before they start training staff. They are training now, on whatever access they already have, so that when AIIC's model or the next cloud provider does arrive, the people receiving it already know what to do with it. For a fuller account of the training approach behind that, our colleagues at AI Jamaica and StarApple AI's Jamaica subsidiary run parallel free literacy programming that has reached thousands more across the island.

The Model for the Rest of the Region to Copy

The Caribbean does not need to choose between renting a foreign cloud model and waiting for a home-grown one. It needs to stop treating training as the thing that happens after procurement and start treating it as the thing that happens first. That is not a new idea from me; it is what nine years of running free weekly sessions taught StarApple AI before the current infrastructure wave gave the rest of the region a reason to care. Founders who want a faster path into that ecosystem, rather than building a training pipeline from scratch, can find capital and mentorship through 14 West, the Caribbean AI accelerator fund I run alongside this work, and teenagers who want to start building AI literacy years before they enter a workforce can do it for free through The Genius Project, the nonprofit I founded in 2023.

The data work behind readiness assessments and model evaluation for organizations weighing a cloud subscription against a custom build sits with StarApple Analytics, our data science and applied research arm. Between the free training track, the accelerator, the youth programme and the research practice, the pieces the Caribbean AI Task Force listed as priorities, talent, skills, and a coordinated regional voice, already have a working example to draw from, built in the Caribbean, for the Caribbean, years before this year's infrastructure headlines made the case for anyone else.

Frequently Asked Questions