Data Centers & Digital Infrastructure

What the segment covers

The Data Centers & Digital Infrastructure segment covers the physical layer of computing rather than the computing itself. It spans colocation and wholesale data center operators, hyperscale self-build campuses, edge and enterprise facilities, and the enabling systems that make them operable — power distribution, cooling, backup generation, on-site storage, interconnection and the fiber routes that tie a facility to the network.

The segment is deliberately defined around capacity rather than software. A data center is a capital asset whose output is measured in megawatts of commissioned IT load, and whose revenue is a function of how much of that load is leased, at what density, and with what redundancy. Everything analytically interesting about the segment follows from that: it is a real-estate and power business wearing a technology label.

Why the segment behaves differently from the rest of IT

Software segments scale on demand. Data centers do not. A facility has a fixed power envelope agreed with a utility, a fixed grid connection date agreed with a system operator, and a construction cycle measured in quarters. Capacity cannot be conjured in response to a demand signal; it must be committed to years in advance, against a forecast, and financed before a single tenant signs.

Three constraints dominate:

• Grid access. The binding constraint in most markets is not land, capital or equipment but a connection agreement and, behind it, transmission capacity to move power to where the load sits.
• Lead-time asymmetry. A data center can be built in roughly 18 months to two years. A transmission project takes about seven. The mismatch is structural and cannot be financed away by the data center operator alone.
• Irreversibility. A shell and a substation are sunk. Operators therefore pre-lease, pre-sell and secure power before committing capital, which makes the segment unusually sensitive to contract quality and unusually exposed to demand forecasts that do not materialise.

Structural characteristics

• High capital intensity, with capex per megawatt of commissioned IT load in the millions of US dollars
• Long asset lives and long tenant contracts, typically five to fifteen years for wholesale and hyperscale
• Concentrated ownership at the operator layer, with global private capital and infrastructure funds dominant
• Bilateral, non-transparent pricing negotiated per deal rather than published per unit
• Regulatory exposure that runs through energy and licensing rather than through telecoms
• Strong locational determinism: capacity follows substations, fiber landing points and latency-tolerant land, not where demand conceptually is

How the segment is evolving

Artificial intelligence has changed the unit of demand. Where a large enterprise facility once measured its requirement in single-digit megawatts, AI training and inference campuses are now specified in hundreds of megawatts, and the largest announced projects in emerging markets run to multiple gigawatts. That shift has three consequences: power procurement has moved from a facilities function to a board-level strategic one; operators are increasingly contracting directly with generators and building or co-owning transmission; and sites are being chosen next to generation rather than next to cities, including at renewable complexes where power would otherwise be curtailed.

The result is a segment in which the scarce input is no longer capital or equipment but an energised connection — and in which the quality of a market can be read from the length and composition of its grid connection queue far more reliably than from its announced investment totals.