Building a Resilient AI Data Centre Strategy for Guyana

Nov 30, 2025 Features / Columnists, News

---

By Karen Abrams, MBA, AA, Doctoral Candidate

(Kaieteur News) – During my years in operations risk management for a major US internet service provider, one principle guided our every decision; redundancy protects nations, not just networks. As Director of Operations Risk Management, I led a team responsible for business continuity planning across three geographically dispersed data centres. They were designed so that if one went down, the others would immediately take over. That was standard practice because once infrastructure becomes mission-critical, failure is not just inconvenient; it is damaging.

Our data centres hosted email for five million users, web hosting for thousands of e-commerce sites, storage and backup services, network and security services, and colocation for enterprise clients. Any outage in that environment could cost our customers millions of dollars each day.

I return to that lesson as Guyana now embarks on an ambitious proposal for a US$2B-plus, 100-megawatt AI data centre, one of the largest infrastructure bets in our modern history. The vision is bold and it pushes Guyana into the global conversation on digital sovereignty and advanced computing. But a project of this scale also demands a clear, technically grounded blueprint for how it will deliver real returns and how it will stand up under stress.

A 100MW facility is enormous by global standards. Inside such a data centre you would expect GPU and CPU server farms, massive data storage systems, high-speed networking, security and governance systems, significant power and cooling infrastructure, platform and application layers, and dedicated operations and monitoring systems. It would also hold the digitized data and security infrastructure needed to run our ministries efficiently and to manage security cameras across the nation.

For context, industry estimates show that 1MW of data centre power can support roughly 1,000 average US homes. A 100MW AI centre therefore draws about the same electricity as 80,000 to 100,000 households (U.S. Energy Information Administration, 2024). Cooling requirements are equally significant; traditional water-cooled data centres often use between 3 and 5 million gallons of (pristine) water per day at this scale (Mytton, 2021; Patterson et al., 2021). Even with more efficient cooling, resource demands remain substantial.

I applaud Guyana for aspiring to build technology capacity. A country that trains and deploys its own AI models (AI sovereignty) holds more control over governmental digitization, national records, education platforms, and cybersecurity. Still, even with full digitization of our public sector, expanded e-services, electronic medical records, and private-sector growth, domestic demand would likely use only a fraction of 100MW. That suggests the business model must rely on hosting regional or international workloads.

This is where the redundancy question becomes unavoidable.

Countries that host foreign data inherit a duty that goes beyond their borders. If Guyana aims to become a digital vault for Caribbean partners, then maintaining uninterrupted services becomes a national obligation. Single-site AI data centres carry concentrated risk. In 2023, an electrical incident at a Microsoft Azure data centre in Australia took down services across parts of the country (Microsoft Incident Report, 2023). Similarly, a fire at France’s OVHcloud data centre in 2021 disabled millions of websites, government portals, and financial platforms (CNRS, 2022). These are cautionary examples of what can occur when redundancy is inadequate.

A distributed model of several smaller data centres is the global best practice for resilience. The US, Ireland, Singapore, and South Korea all rely on multi-site architectures so that if one centre experiences a disruption, others assume the load (OECD Digital Economy Papers, 2023). Distributed systems are also more adaptive. They reduce stress on national grids, support edge computing (allowing computing to happen closer to where people live and work), and scale in alignment with real demand.

Financial planning must also account for recurring capital requirements. AI data centres rely on specialised and exorbitantly expensive GPUs such as NVIDIA’s H100 or its successors. These chips are on rapid upgrade cycles, often every 24–36 months (NVIDIA Investor Briefing, 2024). If a facility is intended to serve international clients, staying competitive means continuously reinvesting hundreds of millions in updated hardware, cybersecurity and maintenance.

None of this diminishes the value of the ambition of our nation’s leaders. Guyana is right to build technological capability. Nations that prepare early for AI transformation become regional leaders. My thoughts are to ensure that our nation’s considerable investment matches global standards for resilience, for the decades ahead.

With a comprehensive plan for redundancy, clear sustainability targets, and transparent long-term financing, this project could anchor a new era of digital strength for Guyana. It could elevate public services, attract high-value technological industries, and signal to the world that Guyana is ready to lead in the future economy.

---

• AI, Data Centre, resilient, US internet service provider

---