As the digital economy expands, data centres are increasingly vital. Tech giants like Microsoft and investment firms such as BlackRock are investing heavily in their construction, recognising their role in future technological advancements. With the rapid growth of artificial intelligence (AI), the demand for powerful data centres
has surged, exemplified by a new 1 GW hydrogen-powered data centre near
Houston, built to support large scale AI operations.
Data centres are the backbone of the digital economy, powering AI, big data, eCommerce, cloud services and more. And, as data processing grows, these facilities require constant upgrades to handle AI’s computational demands. Despite the enormous financial and strategic investment needed to build and maintain them, their importance continues to rise as AI and machine learning reshape industries.
Data centres enable essential AI functions, including real-time medical imaging, autonomous vehicles, and low-latency applications. Their design varies depending on use cases; for instance, crypto mining requires cheap power, while AI workloads prioritise computing resources and minimal latency. As digital transformation accelerates, businesses rely on data centres for cloud applications, Enterprise Resource Planning (ERP), Customer Relationship Management (CRM) and collaboration tools.
CORE COMPONENTS OF AI DATA CENTRES
AI workloads involve three key phases: data preparation, AI training, and AI inference. Data preparation focuses on collecting and curating large datasets. AI training is the most resource-intensive phase, requiring powerful GPU clusters to process vast amounts of data
iteratively. AI inference applies the trained model to real-world scenarios for decision-making.
AI training demands significant computational resources and robust networking infrastructure. GPU clusters, sometimes involving tens of thousands of servers, must communicate without bottlenecks to optimise performance and reduce tail latency, which occurs when slow outlier workloads delay overall job completion. Given that high-end AI servers can cost over $400,000 each, ensuring 100% network reliability is
essential to maximise return on investment.
While GPUs drive AI computing power, networking efficiency determines their full potential. The latest Ethernet-based AI data centre networks incorporate congestion management to prevent traffic slowdowns, load balancing to distribute AI workloads efficiently, and latency
minimisation to ensure fast job completion times. Automation and simplified management also play a key role in maintaining long-term performance and reliability. Historically, InfiniBand was the dominant technology in high-performance computing and AI training networks due to its speed and efficiency in transferring data. However, Ethernet is now
emerging as the preferred alternative, offering greater scalability, lower costs, and a larger talent pool for management and optimisation.
Beyond networking, an AI-ready data centre consists of network infrastructure connecting servers, storage, and external data sources. Storage systems hold vast AI datasets, while computing resources provide processing power through GPUs and AI-optimised CPUs.
AI data centre networking fabric is critical to ensuring that even the most advanced hardware can efficiently feed data into AI algorithms. Without it, performance and scalability would be severely limited, making high-level AI processing inefficient and costly.
THE GLOBAL DATA CENTRE LANDSCAPE
As of March 2024, more than 11,800 operational data centres exist worldwide. The United States leads with 5,388 facilities, accounting for 45% of the global total. Germany (520), the United Kingdom (512), China (449), and Canada (336) follow as the next largest markets. While these numbers highlight the distribution of data centres, variations in size
and capacity mean that some facilities hold significantly more computing power than others.
Africa Data Centres is the leading operator of data centres across the continent, with major hubs in South Africa, Kenya, Nigeria, and other key regions. It is part of the Cassava Technologies group and offers interconnected, carrier- and cloud-neutral facilities. The company focuses on ensuring connectivity by linking its centres to major internet exchanges like Johannesburg Internet Exchange (JINX) and Cape Town Internet Exchange (CINX). While South Africa remains the most developed market, Nigeria and Kenya have seen rapid growth, with emerging hubs in Egypt and Morocco.
South Africa hosts the continent’s largest data centres, with its top ten facilities boasting a combined IT load of 258 MW and total whitespace exceeding 112 000 square metres. Whitespace, representing the available space for servers, and IT load, indicating power capacity, are key metrics used to compare data centres. The ranking of these centres
prioritises whitespace over IT load, as larger facilities do not always equate to higher power capacity. Cloud providers like Amazon Web Services (AWS) are excluded from the rankings as they do not disclose their data centre sizes.
Larger facilities include NTT Data’s Johannesburg 1, which provides 6 200 square metres of whitespace and a 12 MW IT load, and Africa Data Centres’ JHB1 in Midrand, which has 8 500 square metres of space and supports up to 20 MW of IT load. The Vantage Johannesburg I facility is set for major expansion, with plans to reach 60 000 square metres of whitespace and 80 MW IT load capacity. Meanwhile, Teraco’s Brackenfell
campus in Cape Town boasts 18 000 square metres of whitespace and a 50 MW IT load, ensuring redundancy in power and cooling systems.
Africa Data Centres is investing in sustainability, recently signing a 20-year Power Purchase Agreement (PPA) with Distributed Power Africa to supply 12 MW of renewable solar energy to its South African facilities. This move aligns with the company’s goal to power all its data centres with clean energy. The agreement contributes to the United Nations
Sustainable Development Goals by ensuring sustainable energy access and
addressing climate change.
WHY AI AND DATA CENTRES ARE INEXTRICABLY LINKED
The rapid expansion of AI workloads is dramatically reshaping data centre demand, particularly in Europe, where power consumption for data centres is expected to nearly triple by 2030. McKinsey projects that European data centre demand will rise from 10 GW to 35 GW, requiring investments between $250 billion and $300 billion. Additionally, energy
consumption is expected to surge from 62 TWh today to over 150 TWh by 2030,accounting for approximately 5% of total European electricity use. However, challenges such as power availability, sustainability concerns, infrastructure constraints, and long lead times for electrical equipment pose significant hurdles to scaling data centres.
Globally, data centre demand is projected to grow at an annual rate of
19-22%, potentially reaching 171-219 GW by 2030. In a more aggressive
scenario, this demand could rise by 27% to 298 GW, creating a potential
supply shortfall. McKinsey estimates that AI and analytics could generate $10 trillion in economic value, but unlocking even a fraction of this potential will require substantial data centre expansion. Around 70% of future demand will be for AI-ready facilities equipped to handle advanced machine learning workloads, accelerating the shift in how and where data centres are built.
AI-driven demand is reshaping data centre infrastructure, pushing both the U.S. and Europe to expand capacity rapidly. AI workloads require immense computing power, driving the need for AI-ready data centres. By 2030, global data centre capacity demand is expected to triple, with AI-specific infrastructure accounting for 70% of that growth. Generative AI alone will represent 40% of total demand.
Goldman Sachs estimates AI will add approximately 200 TWh per year to global data centre electricity demand between 2023 and 2030. While AI innovations may improve computational efficiency, power consumption is still expected to increase significantly. By 2028, AI will account for around 19% of total data centre electricity use. Addressing this rapid expansion presents both an infrastructure challenge and a major
investment opportunity across the power and data centre sectors.
THE DATA CENTRE ECOSYSTEM
The AI Data centre ecosystem framework provides a structured approach
for integrating AI into data centre operations, optimising infrastructure, software, and processes. This framework ensures scalability, security, and efficiency, helping organisations process data faster, improve energy use, and maintain high security standards.
A critical component of AI-ready data centres is networking infrastructure, which enables the smooth flow of data into AI algorithms. Unlike traditional data centres that primarily store information, AI workloads require high-bandwidth, low-latency networking to efficiently handle massive parallel processing. Without this fabric, data movement becomes a bottleneck, hindering AI performance.
Key characteristics of AI data centres include high-density deployments requiring up to 50 kW per rack and specialised designs, along with advanced liquid cooling solutions to manage the heat from AI workloads.
They demand robust networking fabrics for high-speed interconnects and superior data flow management. Operations are increasingly powered by AI-driven management tools for enhanced efficiency and automation, and AI also helps address workforce shortages by automating maintenance and monitoring tasks.
The AI-driven data centre ecosystem ensures that physical and digital infrastructures work together seamlessly. Without a well-structured networking fabric, AI models will struggle to access and process data effectively, limiting their potential impact.
BUSINESSES ARE EMBRACING AI DATA CENTRES
Many leading companies are investing heavily in AI data centres to support their operations. Google has committed billions to AI data centres to enhance its AI products, such as Google Search and Google Assistant, utilising custom TPUs for workload acceleration. Nvidia is developing cutting-edge architectures for GPUs, DPUs, and CPUs,
including the Blackwell GPU, which introduces six transformative technologies for AI processing, training, and deployment. Tesla relies on AI-driven data centres to power its autonomous vehicle technology, using AI algorithms to process data and make real-time decisions. Alibaba has built a vast network of AI data centres to sustain its e-commerce services and plans to develop new, environmentally friendly facilities.
While industries like technology and automotive have embraced AI data centres, the healthcare sector has been more hesitant due to concerns over compliance and data sovereignty. However, as AI continues to demonstrate tangible benefits, such as improved security measures, it is expected that more healthcare organisations may adopt AI-driven data centre solutions.
The growing reliance on AI-powered data centres presents both challenges and opportunities. While the energy demands of AI are increasing, innovative solutions can help make data centres more efficient and sustainable. Advances in cooling technology, such as
purpose-built cooling systems, significantly reduce energy consumption. Additionally, AI-driven optimisation tools allow data centres to dynamically adjust operations, further enhancing efficiency while maintaining reliability. Through strategic improvements in energy management and infrastructure, AI data centres can continue to evolve,
supporting the increasing global demand for AI applications while minimising their environmental impact.
The integration of AI into data centres presents significant challenges in terms of readiness, infrastructure, and workforce. The increasing demand for power, the shift to decentralised infrastructure, and the shortage of skilled talent require urgent attention for businesses to
keep up with evolving data needs.
