Table of contents
The global data centre industry is approaching a structural turning point. According to JLL’s 2026 Global Data Centre Outlook, nearly 100 gigawatts of new capacity could come online between 2026 and 2030, effectively doubling global supply in five years (JLL, 2026). Forecasts indicate compound annual growth of around 14% through 2030, driven primarily by AI and cloud workloads (JLL, 2026).
This is no longer a specialised real estate segment. Data centres are becoming the physical backbone of the artificial intelligence economy. If projections are realised, the transformation will not be cyclical but structural: digital infrastructure will evolve from an enabling layer to a strategic foundation.
From the experimental phase to permanent infrastructure demand
In recent years, the growth of AI infrastructure has been driven primarily by clusters dedicated to training large models. This dynamic is changing. By 2030, AI workloads could account for about half of total data centre demand, with inference set to surpass training as early as 2027 (JLL, 2026).
The economic difference is significant. Training is episodic; inference is persistent. Once integrated into business workflows, customer interfaces, supply chains, and financial systems, models generate continuous computational demand. As already highlighted in previous analyses on the full-stack evolution of AI, value tends to migrate towards those who control the infrastructure layers that enable scalability and reliability, not just towards those who develop applications (Agostini, 2025a).
This transition transforms AI from experimental software to utility-scale infrastructure consumption.
Capital and concentration, the new competitive logic
The potential investment requirement is significant. JLL estimates up to three trillion dollars between real estate and IT by 2030 to support the expected growth (JLL, 2026). Deloitte (2025) and IoT Analytics (2025) highlight how hyperscalers are allocating hundreds of billions a year to AI infrastructure.
This level of capital intensity changes the competitive landscape. High financial requirements favour operators with access to patient capital, large-scale procurement capabilities and industrial execution skills. Data centres are increasingly treated as long-term core infrastructure assets, rather than speculative technology bets (JLL, 2026).
The transition is strategic: the logic of software multiples is giving way to a logic of capital deepening. AI is entering a phase of industrial buildout.
Energy, an enabling variable and systemic constraint
Among all the variables at play, electricity emerges as the most immediate factor. The International Energy Agency predicts that data centres could exceed 500 terawatt hours of global consumption as early as 2026, equivalent to approximately 2% of global electricity use (IEA, 2026). In some markets, the growth in demand from data centres is approaching planned new generation capacity, signalling potential grid strains (BloombergNEF, 2025). In the United States, data centre consumption could exceed 4% of the national total by 2030 (Pew Research Centre, 2025).
This means that access to energy is becoming the gating variable for AI expansion. It is no longer just a question of chips or models. It is a question of the electricity grid, authorisations, generation capacity and flexibility.
Hyperscalers are responding through renewable procurement, behind-the-meter generation and storage system integration. The evolution of renewable costs and grid flexibility planning will be key to sustaining growth (IEA, 2026; BloombergNEF, 2025).
Governance and geopolitics: friction or catalyst?
The trajectory of the supercycle is not inevitable. It depends on a set of interdependent variables: enterprise adoption of AI, capital market liquidity, progress in semiconductor efficiency, and regulatory stability.
More and more governments are treating AI infrastructure as critical national infrastructure, influencing location, sustainability standards and alignment with industrial policies (JLL, 2026). Governance can act as a drag, slowing down deployment through compliance requirements, or as a catalyst, accelerating buildout through strategic priorities and public-private coordination.
As already noted in analyses of the relationship between governance and competitiveness in the agentic economy, regulatory architecture is not limited to containing risk: it can consolidate competitive advantages when it is consistent with infrastructure strategy (Agostini, 2025d).
Systemic dynamics, reinforcement and balancing cycles
The expansion of AI infrastructure can be interpreted as the interaction between reinforcement cycles and balancing cycles.
Greater AI adoption generates greater demand for inference. Demand requires new infrastructure capacity. Increased capacity reduces latency and increases reliability, encouraging further adoption. This reinforcing cycle drives towards concentration and accelerated buildout.
At the same time, energy constraints, financing conditions and social resistance act as balancing forces. Infrastructure does not scale in a vacuum, but within physical, financial and political limits.
The supercycle will only be sustainable if the balancing forces evolve in proportion to the speed of the strengthening cycles.
The strategic issue for leaders
For senior executives, the central question is no longer just about the power of AI models. It is about positioning within the energy, financial and regulatory systems that enable scalability.
Ensuring flexible energy access, structuring resilient capital stacks, designing facilities that are adaptable to computational density, and integrating governance from the earliest stages of development are not marginal operational decisions. They are strategic levers.
AI is moving from the application layer to the infrastructure layer (Agostini, 2025a). Energy acts as both an enabler and a constraint (IEA, 2026). Capital markets increasingly treat digital infrastructure as a structural allocation (JLL, 2026). Governance helps define the architecture of the sector, not just regulate it (Agostini, 2025d).
The crucial question is not how powerful AI systems can become. It is whether the physical and institutional systems surrounding them will be able to scale at the same speed. (photo by Tanner Boriack on Unsplash)
References
Agostini, M. (2025a). AI’s full-stack moment: Why investors must look beyond the model. Medium.
Agostini, M. (2025d). From compliance to competitiveness: How governance and data will define the agentic economy. Medium.
BloombergNEF. (2025). AI and the power grid.
Deloitte. (2025). AI-driven data center infrastructure outlook.
International Energy Agency. (2026). Electricity 2026.
IoT Analytics. (2025). Data center infrastructure market report 2025.
JLL. (2026). 2026 global data center outlook.
Pew Research Center. (2025). Energy use at U.S. data centers amid the AI boom.
S&P Global. (2025). Data center grid power demand to rise.
ALL RIGHTS RESERVED ©