Every AI and energy infrastructure investment is built on commitments: to capacity, to timelines, to jobs, to water and to the communities that host it. The Loregnard Group works on both sides of those commitments, testing whether they can be delivered before they are made, and building the structure to honor them once they are.
Announcements come with capacity figures, timelines and pledges on jobs, water and community benefit. What is harder to find is the work behind them: evidence that the location can deliver what was promised, and the operational structure that carries each commitment once it is made, with clear ownership and a record that is ready when a community, a regulator or an investor asks for it.
That distance does not surface gradually. It arrives at the moment it is most costly. The ones who are not caught in that moment made a different choice earlier.
A public commitment used to be a matter of reputation. An operator made a pledge on water stewardship or community impact, and what it protected was the story it told about itself.
The capital changed that. Financing now arrives with sustainability targets written into its terms, and when a target is missed, the interest rate moves. What was once a difficult news cycle is now a more expensive loan, carried for as long as the debt lasts.
That changes who is asking, and what they can ask for. A regulator asks for the record in the form the statute specifies, not the form that was easiest to produce. A community asks what changed on the ground, and holds the answer against what it was promised. An investor asks whether the target is being met on schedule, verified by someone other than the party reporting it. The Loregnard Group builds the structure that answers all three without contradiction.
The Loregnard Group tests a proposed AI infrastructure investment against the evidence for delivery: what is already in place, what remains conditional, who controls the dependencies, and whether it can all come together in time.
20 MW can be delivered on schedule. The full 30 MW depends on two items that are not yet in anyone's timeline.
Power, land, connectivity, workforce programs and public support may all exist, and the conditions a particular investment needs can still fail to come together on the timeline it requires.
Readiness is whether what an investment depends on can be delivered, at the scale it needs, when it needs it.
Power, connectivity, site and operating requirements, and what has to be resolved before stated capacity becomes usable capacity.
The roles needed to build, commission and run the asset, which of them exist locally, and which can realistically be developed in time.
The approvals, responsibilities and sequencing that sit between what a location says it can do and what it can deliver on schedule.
Workforce, local benefit and community commitments, tested against the real conditions of the investment, with a clear way to show later whether they were kept.
Engineering, utility, environmental and legal determinations stay with the qualified parties responsible for them.
The Loregnard Group brings those findings into one view for the investment decision: what the evidence supports, what remains conditional, how the dependencies affect each other, and whether it can all come together in time.
You do not need to know whether the location is ready. The assessment is designed to determine what can be supported, what depends on other actions, and what needs to happen next.
The methodology is consistent. The context changes. The accountability distance between public commitment and operational reality shows up the same way whether the operator is building a data center or a power plant, whether the commitment was made in a public pledge or written into the terms of the financing.
The accountability conversation is happening on conference stages across multiple continents. The Loregnard Group participates in that conversation actively, bringing the same rigor to a public stage that it brings to a client engagement.
The Loregnard Group was built out of a pattern observed across more than twenty years of building execution infrastructure inside complex, high-stakes environments: commitments made at the leadership level stall before they land, not from carelessness, but because the operational architecture to carry them was never designed. The methodology behind this work was developed and stress-tested across big tech, public higher education, and one of the largest urban school systems in the world.
AI infrastructure accountability is where that work is most consequential right now. The Loregnard Group works with AI data center operators and energy developers building at scale in communities where that distance is creating compounding risk. Engagements outside those practice areas are considered on a case by case basis.
In the race for AI infrastructure, a location's speed to usable capacity is starting to carry real weight alongside its scale. In September 2026, CNBC reported that Anthropic and OpenAI were looking for 20 to 30 megawatt (MW) data center deployments in the UK and the Nordics, and while nothing had been announced, the discussions were a useful example of why.
Much of the data center conversation has been about size: how much power a market has, how large a site can get and whether the grid can carry what AI is asking of it. Training new models still requires enormous, tightly connected computing clusters. But as more people and businesses use AI every day, inference, the computing that happens each time someone uses a model that has already been trained, is creating another kind of demand.
Those millions of requests do not all need to be handled in one place. Twenty MW that can come online sooner in one location and 30 MW in another can each carry a real share of the load while much larger campuses take longer to develop.
That shifts the question for any country or region weighing AI infrastructure investment. If buyers are willing to take capacity across multiple locations, a market does not need hundreds of MW ready for a single development to be part of the conversation. But having 20 or 30 MW somewhere on the grid is not the same as being ready for a buyer.
An advantage goes to markets that can show what they can deliver rather than simply what they have. That means knowing which sites are viable, how much power can reach them and when, how reliable it is, and what connectivity, permitting and operating capacity are already in place. Getting a location to the point where it can answer those questions with confidence is its own buildout, and it is the work TLG does.
Time has always mattered in this industry, but the ability to add capacity across multiple locations gives it more weight. Twenty MW that can be used sooner may be worth more to a buyer than 200 MW that remains years away. That only works in a market's favor if the capacity is actually available when the buyer needs it, which is why the more useful measure of infrastructure readiness is not simply what a location has, but what it can deliver for a particular investment, and by when.
A country can be ready for a data center investment without being ready for what the data center requires over its lifetime.
I've been thinking about that distinction as I follow AI infrastructure development across the U.S., Europe, Africa and the Caribbean. Readiness is still too often measured by what can be secured: land, power, water, permits and capital. That tells us whether a project can move forward. It does not tell us whether the market can support and absorb the investment over time.
We have seen the consequences in mature markets. Grid constraints around Dublin led Ireland to restrict new data center connections. New York has worked through its own version, pausing new hyperscale development while it addresses energy demand, water, ratepayer exposure and community benefit. In both cases the lesson is the same: attracting infrastructure and being prepared to sustain it are different propositions, and recalibrating becomes far harder once development outpaces the systems meant to support it.
For African and Caribbean markets negotiating the next wave, the opportunity lies less in caution than in being deliberate about what readiness means before development reaches scale.
Nigeria offers a precedent from a different sector. Its oil and gas local-content regime spent years building mechanisms to assess in-country capacity before work is sourced externally, and the country is now applying that instinct as it pursues cloud, data center and AI infrastructure alongside stronger indigenous capability. The sectors differ, but the principle holds: establish capability before negotiating the commitment.
For data centers, a capability baseline extends well beyond available workers. It means understanding which roles disappear after construction and which remain for the life of the asset, what technical capability exists and what can be developed, and whether the power, water, training systems and institutions surrounding the asset can sustain what is being proposed.
That baseline cannot simply travel from one market to another. South Africa is not Nigeria. New York is not Dublin. The resource profile, technical base, training capacity and negotiating position all change with the market.
Country readiness needs a baseline before it needs a target.
If a country is committing finite resources to position itself in the AI economy, the return cannot be measured only by whether the facility was built. The better question is what the country is positioned to retain from it twenty years later.
Power matters. Water matters. Jobs matter. But so do the capabilities, institutions and technical knowledge that remain long after construction ends. That is where I think the real work of AI infrastructure sits.
New York became the first state to require large data center developers to fund host-community benefits and meet apprenticeship and workforce-training obligations. Towns that once had little leverage now have a framework that tells them what to ask for.
A framework that says what to ask for does not say how to confirm you received it. A town secures a local hiring commitment, the facility opens, the people who negotiated it move on, and nobody is left checking whether that number was ever met.
The tools to check it already exist. Prevailing wage work produces certified payroll, apprenticeships are registered with the state, and between them they show whether a hiring commitment was actually honored. What no one has required is that those records get read against the promise, on a schedule, in a form the community can use.
That requirement is worth building in now, while the framework is still being written. Before signing, verification is a condition of the deal; after, it is a request the developer can ignore.
So the question a host community should be asking is not what are we owed. It is how will we know we received it.
Kenya launched sovereign cloud platforms in 2026. Nigeria, Ghana, and South Africa have live platforms. Across the Caribbean, governments and banks can now keep citizen records and core banking inside the region rather than on a server another nation's laws can reach.
After years of insisting that where data lives matters, the argument is finally being heard.
But arriving infrastructure and arrived sovereignty are not the same thing. A sovereign cloud operated under foreign contracts, maintained by flown-in expertise, and secured by talent trained on someone else's terms means you own the building and rent the know-how to run it.
That is why workforce is not an education topic sitting politely beside the infrastructure conversation. It is the infrastructure conversation. The training pipelines forming across the Caribbean and Africa are largely designed by the platforms arriving with the infrastructure rather than by the nations hosting it. Whoever shows up with the program ends up setting the terms of the skills.
A nation that intends to own its digital future has to make workforce a negotiated term rather than an inherited one. When the infrastructure agreement gets signed, workforce belongs inside it with numbers, dates, and a name attached to delivering them. Anything less will not survive the first change of government or the first budget cut.
Barbados shows where this stands. Workforce written directly into the national roadmap, a deliberate decision, made and owned, while the curriculum underneath it still belongs to a foreign platform. The will is arriving before the architecture.
Sovereignty is not just where the data sits. It is who runs the infrastructure, who trained them, and who decided what that training would be.
If you have ever asked AI to help you with something you probably would have paid someone else to handle, welcome to my world. I had AI help me figure out my toddler's sleep schedule. I used it to build out a workout routine. In my mind I was saving money. Then one day I asked myself a different question. Not about my subscription cost. About the hidden one.
What I found was not in the headlines we are used to seeing about AI. It was something older and more fundamental. It was water.
A typical large data center uses up to 5 million gallons of water per day, the equivalent of the daily water supply of a town of 50,000 people. Google consumed approximately 5.6 billion gallons of water in 2023, a 24 percent increase from the prior year, driven largely by AI infrastructure expansion. The International Energy Agency identified water as a critical and underregulated input in AI infrastructure globally.
The data on AI water consumption is not classified. It is in annual sustainability reports, in permit filings, covered by Bloomberg, Brookings, and the International Energy Agency. It has been there. But it was never designed to reach us. The conversation exists in fragments but none of it has reached the everyday user in a way that is coherent, personal, or actionable.
Communities have always paid the bill for infrastructure decisions made above them. AI infrastructure is not creating this pattern. It is funding it at scale and at speed. The bill is always paid by the ones who cannot afford it. Until suddenly the bill is large enough that nobody can afford it. And by then the infrastructure is already in the ground.
Two things can be true at the same time. AI is useful. It has changed how I work, how I access information, how I navigate systems that were not built for people like me. I am not here to argue against it. I am here to argue that the benefits we are experiencing should not come at the exclusive expense of communities that had no seat at the table when the decisions were made. When I look at my son I know he will grow up in a world I cannot fully fathom. The price of progress is high. My son will not have the luxury of not knowing that price. Let this be your catalyst.
Every organization has a version of this meeting. The plan gets approved, the next steps go into the system, and someone has already updated the tracker before they have even left the room. Pull up the dashboard on your way out and everything looks exactly like progress is supposed to look.
What I am describing is not bad decision-making. It is incomplete decision-making, which is a harder problem to name because it looks fine from the outside.
The person who flagged a concern in the pre-read got a polite redirect rather than a real answer. The ownership question that never quite resolved became someone's problem to figure out alone, quietly, while the timeline kept moving. The plan that got approved was the plan the room could say yes to, not necessarily the plan the work required.
Dashboard movement is not the same thing as initiative viability. Checking things off, advancing tasks, keeping the status green tells you nothing about whether the conditions that have to be true for this to actually land were ever examined. An organization can be fully in motion on something that was never going to work the way it was described, because the approval happened before anyone asked whether the sequencing was real, whether the ownership was genuine, or whether the timeline reflected what the work actually required. By the time that becomes visible, the conversation has shifted from whether it will work to who is responsible for the fact that it did not. That conversation is expensive in every direction and in most cases it was avoidable.
The point of leverage is the moment before the decision gets made, when there is still room to ask what has to be true for this to work, and what it costs the organization if any of those conditions turn out to be wrong. Not at the task level. At the level of what the organization is genuinely willing to trade in time, in capacity, in trust, if this goes the way underprepared initiatives tend to go. That question almost never makes the agenda. The discipline to pause and ask whether what is being approved actually reflects what was decided is the part that consistently gets left behind.
A dashboard shows you what was approved. It has nothing to say about whether what was approved was fully thought through. That work happens before anyone opens the system, in the room where the decision is still forming. The organizations that run that room well do not look like they are moving slower. They look like they finish what they start.
Every engagement starts the same way: a conversation about scope, runway, and timeline. What you are working with, where the distance lives, and what closing it looks like before someone else defines the terms.
The Loregnard Group works with governments, developers, operators and investors in the United States, the Caribbean, and Africa. We look forward to starting a conversation.
Discuss an AssessmentThe Loregnard Group Corp is a US-based advisory firm. This policy describes how we collect, use, and handle information when you visit theloregnardgroup.com or contact us directly.
When you visit our website, we may receive basic technical data standard to web hosting. If you submit your email address, we retain it to send occasional perspectives and updates. If you contact us directly by email, we retain that correspondence to respond to your inquiry and, where applicable, to fulfill an engagement.
TLG uses AI-assisted tools, including large language model platforms, to support research, drafting, analysis, and internal workflow. We do not input identifiable client data or confidential engagement information into third-party AI tools without appropriate safeguards. AI is part of how we work. It is not a replacement for our judgment.
We use information received to understand how our site is accessed, to respond to inquiries, to send subscribers occasional perspectives, and to deliver advisory services. We do not sell, trade, or share visitor information with third parties for marketing purposes.
Questions may be directed to: denessa@theloregnardgroup.com