For most of the past two decades, data centers were treated as a back-office category of real estate — important, but rarely discussed in the same breath as ports, power plants, or rail networks. That framing is changing quickly. The capital intensity, construction scale, and economic centrality of AI infrastructure now place it firmly alongside the industrial assets that define a nation's productive capacity.
A Capital Cycle Without Precedent in the Digital Economy
The scale of investment is the clearest signal. According to the International Energy Agency, the capital expenditure of just five large technology companies exceeded USD 400 billion in 2025 and is projected to grow by a further 75% in 2026. That capital is not being spent on software licences or marketing — it is being spent on land, concrete, steel, transformers, cooling plant, and silicon, deployed at a pace and scale that increasingly resembles heavy industry rather than conventional IT procurement.
The IEA's satellite-based tracking shows that "AI factories" — facilities purpose-built for AI training and inference — have more than tripled in capacity in the past eighteen months alone. This is not incremental growth within an existing asset class. It is the emergence of a new category of industrial infrastructure, with its own supply chains, siting logic, and risk profile.
Electricity Demand Is the Tell
Perhaps the most telling indicator that AI infrastructure has become industrial-grade is its relationship with electricity systems. Global electricity demand from data centers grew 17% in 2025, broadly in line with IEA projections, while consumption from AI-focused facilities specifically surged by 50% over the same period — far outpacing the 3% growth rate of global electricity demand overall.
- Data center electricity consumption is on a trajectory to roughly double from approximately 485 TWh in 2025 to around 945 TWh by 2030
- AI-focused consumption within that total is expected to triple over the same period
- Grid planners in multiple jurisdictions are now treating data center load growth as a primary input to capacity planning, not a marginal one
No category of conventional commercial real estate moves electricity markets, transmission planning, or industrial policy in this way. Power plants, smelters, and heavy manufacturing do. AI infrastructure is increasingly being planned, financed, and regulated on the same terms.
When a single category of digital infrastructure begins to shape national electricity planning, it has stopped being a real estate asset class and started being industrial infrastructure.
Why This Reframing Matters for Developers and Investors
Treating AI infrastructure as industrial infrastructure changes how projects should be evaluated and delivered. It means power availability, not floor area, becomes the primary site selection constraint. It means electrical equipment lead times — transformers, switchgear, generation — become as critical to underwriting as construction cost. And it means governments increasingly view large AI facilities through an industrial-policy lens: jobs, energy security, and sovereign technological capability, not just tax revenue from a data center campus.
For institutional investors, this reframing also changes the comparison set. AI infrastructure assets increasingly compete for capital and attention with energy infrastructure, utilities, and industrial logistics — sectors with established frameworks for evaluating long-duration capital projects, power purchase structures, and regulatory exposure.
Industrial Infrastructure Comes With Industrial Discipline
One useful test for whether an asset class has matured into genuine industrial infrastructure is whether it has developed the institutional discipline that category demands: long-horizon capital planning, standardised technical due diligence, and a workforce and supply base purpose-built for the sector rather than borrowed from adjacent industries. AI infrastructure is moving through exactly this maturation curve. Engineering, procurement and construction firms are building dedicated data center divisions. Equipment manufacturers — from transformer and switchgear suppliers to cooling specialists — are expanding capacity specifically to serve this demand category rather than treating it as one customer segment among many.
This maturation cuts both ways for market participants. It raises the bar for what counts as credible execution capability, making it harder for under-resourced developers to compete on speed alone. At the same time, it creates a more investable asset class for institutional capital that has historically been reluctant to underwrite bespoke, one-off infrastructure projects — the more AI infrastructure resembles a repeatable industrial process, the more comfortably it fits within conventional infrastructure fund mandates.
The Policy Dimension Is Already Catching Up
Governments are responding to this industrial reframing in real time. Energy ministries, grid regulators, and industrial policy bodies are increasingly treating large AI infrastructure programmes as strategic assets warranting the same coordination historically reserved for energy generation, semiconductor fabrication, or transport infrastructure — including, in some jurisdictions, expedited permitting pathways, targeted grid investment, and direct engagement on siting strategy. This is a meaningful departure from the largely hands-off regulatory posture that characterised earlier generations of data center development, and it is a trend developers and investors should expect to continue rather than reverse.
DATAPERT's Perspective
At DATAPERT, we advise clients to treat AI infrastructure programmes with the same rigour applied to large industrial projects: power-first site selection, early engagement with grid operators and equipment suppliers, and investment structures that explicitly price in interconnection and supply chain risk. Explore how DATAPERT supports clients across data center development and investment strategy, from early feasibility through to delivery.
