The AI gold rush has a dirty secret: most infrastructure investors are backing the wrong operators. As tech giants pour over $320 billion into data centres this year alone – up from $241 billion in 2024 – a critical question divides winners from losers. It’s not who has the most facilities. It’s who can actually scale them.
“Not all infrastructure investments are created equal,” says Nnadozie Odinaka, a strategic finance professional completing his MBA at Georgia Institute of Technology, who has advised business leaders across Africa, Europe and North America on scaling challenges, in an interview with The Guardian. “Investors fixate on occupancy rates and power capacity. The real question is whether your target operator has the balance sheet to scale at the speed AI demands.”
The numbers reveal the stakes. AI workloads consume four to six times more power per server than traditional computing. Advanced GPU-equipped racks now demand up to 500 kilowatts – enough to power a neighbourhood. A single AI training cluster requires 100-plus megawatts, equivalent to a small city’s power grid.
Yet Odinaka, who cut his teeth in financial risk advisory at leading consulting firms, sees investors consistently underestimating two risks: capital intensity and execution complexity.
“If your data centre operator can’t secure power substations, cooling infrastructure, and construction permits in advance, they’ll miss the market window entirely,” he warns. “By the time they’re scrambling for megawatts, you’re already losing.”
Drawing on his Big Four audit and finance experience analysing infrastructure risks, Odinaka identifies the warning signs.
First, insufficient power procurement pipelines. “The operators winning right now secured power contracts two years ago,” he notes. “If they’re just starting negotiations, you’re already behind.”
Second, outdated cooling technology. AI density demands liquid cooling systems most legacy operators lack. Retrofitting costs can exceed 50% of original budgets.
“Cost overruns of 50 to 100 percent are common in this space,” Odinaka says. “Your operator needs financial breathing room.”
For infrastructure investors accustomed to real estate fundamentals, Odinaka recommends a different framework.
“Stop looking at current capacity utilisation – that’s backward-looking,” he advises. “Ask whether they can deploy twice to thrice their current capacity within 18 months. Ask about reserved power capacity for the next five years. Ask if their capital structure can absorb the unexpected.”
The winning operators share three characteristics: strong utility partnerships, modular construction capabilities, and diversified funding sources. “They’re treating scalability as a financial engineering problem, not just a construction problem,” he explains.
Economic research projects AI could add $15 trillion to global GDP by 2030 – but only if infrastructure scales proportionally. For investors, that’s generational wealth creation potential. For the underprepared, it’s stranded assets.
“In five years, we’ll see clearly which operators understood that data centres are the foundation of the AI economy,” Odinaka says. “Foundations determine how high you can build – and how much value you capture.”
His message to the infrastructure investment community is blunt: AI is real, infrastructure demand is real, but not every bet pays off.
“This isn’t about picking the operator with the most facilities,” he concludes. “It’s about identifying who has the financial discipline, strategic foresight, and execution capability to scale profitably. The operators building smarter financial frameworks and securing critical inputs ahead of demand – those are your winners.”
For investors betting on AI’s infrastructure layer, scalability isn’t optional. Neither is the financial sophistication to deliver it.
