$ 7 trillion The data centre investment McKinsey estimates could be required globally by 2030.
SIGNALS: PEOPLE, NUMBERS, AND IDEAS SHAPING BUSINESS
IDEAS in motion
AI’ S REAL BOTTLENECK IS NOT THE CHIP
The artificial-intelligence boom is creating an enormous market for less celebrated technologies: electricity transformers, generators and cooling systems.
Nvidia’ s chips may have become the symbols of the artificial-intelligence economy, but a chip is useful only if somebody can provide the electricity to run it and remove the heat it produces.
The pressure is producing a new group of industrial winners. South Korea’ s HD Hyundai Electric reported an equipment order backlog of $ 8.5 billion at the end of June 2026, covering more than three years of production. China’ s Jinpan Technology saw new data centre orders more than quadruple in the first half of the year.
Cooling presents another opportunity. More powerful AI chips generate much more heat than conventional computing equipment. Bank of America estimates that electricity consumption from a single AI rack could exceed 1.5MW by the end of the decade, almost 100 times the consumption of a conventional rack.
That is becoming increasingly difficult.
McKinsey forecasts that almost $ 7 trillion could be invested in data centres globally by 2030. Developers are racing to increase capacity, but projects are encountering shortages of transformers, generators, cooling equipment and grid connections.
Operators are consequently moving from airconditioned server rooms towards systems that circulate liquid directly around computing equipment. Liquid cooling could account for 70 % of new AI data centre installations by 2030, compared with approximately 30 % today. It can also reduce the energy consumed in cooling by more than a quarter.
Technology companies may want a new facility operational within six months. Connecting it to an electricity grid can take as long as two years in some emerging markets and more than eight years in parts of the developed world.
At the edge of the server hall, the AI economy looks less like software than heavy industry. Orders for those systems are already accumulating faster than some manufacturers can supply them.
The numbers
$ 7 trillion The data centre investment McKinsey estimates could be required globally by 2030.
Why it matters for South Africa
South Africa is estimated to host about 70 % of Africa’ s existing data centre capacity, giving it an early advantage in the contest for AI and cloud investment.
But future growth will place significant demands on electricity, water and municipal infrastructure. Two facilities proposed by Equinix in Cape Town could require approximately 170MW of power, provoking objections over their environmental and community impact.
Eskom’ s improved generating performance gives the country an opportunity to attract new industrial electricity customers. The danger is assuming that available generation automatically means suitable connections, transmission capacity, and sustainable water supplies.
36 sabusinessintegrator. co. za