More than 90% of stakeholders across the data centre value chain have experienced a material infrastructure disruption in the past five years, according to a global survey of 1,800 leaders.
The survey, conducted for "Built for What's Next", a research programme from Economist Enterprise sponsored by insurer FM found that more than 80% of respondents prioritise growth, efficiency and AI deployment over long-term resilience when forced to make trade-offs. The findings come even as roughly two-thirds report board-level oversight of resilience and three-quarters say they invest in targeted resilience initiatives.
Global data centre capacity is projected to nearly triple to 220 gigawatts by 2030, requiring about $6.7 trillion in infrastructure investment, according to the report. Those figures align closely with separate estimates published by McKinsey, which found in a report this year that data centres will require $5.2 trillion in investment by 2030 to meet worldwide demand for artificial intelligence alone, based on a projected 156 gigawatts of AI-related capacity demand. McKinsey's estimate could range from $3.7 trillion to $7.9 trillion depending on how quickly AI adoption accelerates.
Commercial real estate firm JLL, in a separate outlook published in January, projected a narrower trajectory: global capacity nearly doubling from 103 gigawatts to about 200 gigawatts by 2030, requiring up to $3 trillion in investment.
"We're witnessing the most significant transformation in data centre infrastructure since the original cloud migration," said Matt Landek, JLL's global division president for data centres and critical environments.
The Economist Enterprise/FM report identified what it called "amplifier risks" – including supply-chain disruption and climate change – that can spread across interconnected systems. More than two-thirds of respondents ranked supply-chain disruption among their top five systemic risks, but only about half expressed confidence in their ability to manage it. Measures addressing those risks, such as geographic diversification and compound-shock simulations, remain less common.
The report also found gaps in cross-organisational planning. Just 10% of those surveyed maintain structured crisis plans with all critical external partners, while a quarter confine planning internally and reach out to partners only after a disruption occurs, the report said. About a third of respondents have cross-functional crisis teams, and 30% run compound-shock simulations.
Independent reporting has separately documented power-grid constraints affecting data centre construction timetables. Climate-research firm Sightline Climate reported in May that 30% to 50% of the roughly 16 gigawatts of data centre capacity planned for 2026 will be delayed or canceled due to power availability, equipment shortages and community opposition. Separately, Goldman Sachs Research estimated this year that only 60% of next year's scheduled capacity will arrive on time, a figure the bank projected would drop to 50% over the following two years.
Those independent findings track closely with the Economist Enterprise/FM survey's own risk rankings: cyber-attacks, grid constraints and extreme weather dominated perceptions of systemic risk, with more than 75% of each stakeholder group naming all three among the five risks most likely to have system-wide effects. By contrast, fewer than 5% of respondents ranked community opposition among their top five systemic risks, and barely a tenth cited technological obsolescence.
The report also flagged human capital as a resilience constraint. Fewer than half of respondents maintain dedicated talent pipelines or training programmes, and about 60% of data centre operators already struggle to recruit qualified staff, the survey found. Separate industry reporting has identified skilled-labor shortages, including among electricians and engineers needed to install and commission grid equipment, as a constraint on new capacity coming online.
The Economist Enterprise research drew on the 1,800-person survey alongside scenario mapping, expert interviews and an advisory board, though the report did not specify the composition of the advisory board or provide additional methodological detail in the material reviewed.
