AI Infrastructure: The Convergence of Data Centers and Electrification

Abstract

Artificial intelligence is driving one of the largest infrastructure investment cycles since the expansion of the internet, requiring not only advanced semiconductors but also hyperscale data centers, electrical equipment and reliable electricity generation. This paper argues that the AI infrastructure investment opportunity extends beyond computing capacity into the broader electrification of the economy, and evaluates a portfolio spanning both themes: electrical infrastructure, cooling and power systems, data-center real estate, AI networking and electricity generation, complemented by a diversified electrification ETF. The combined thesis emphasizes that electricity availability, not computing hardware, has become the principal constraint on AI deployment.

Key findings

1. 01McKinsey estimates data centers may require approximately $6.7 trillion of cumulative capital expenditure through 2030, with roughly $5.2 trillion tied to AI workloads, while global data-center capacity demand is projected to rise from approximately 82 GW in 2025 to 219–220 GW by 2030.
2. 02The IEA projects global data-center electricity consumption will reach approximately 945 TWh by 2030 — more than double current levels and growing roughly four times faster than electricity demand across the rest of the economy.
3. 03The five core holdings (Eaton, Vertiv, Equinix, Arista Networks, Constellation Energy) combine above-average revenue growth with strong profitability — Arista's operating margin exceeds 40% — though several already trade at premium valuations, with Equinix's P/E near 85x.
4. 04The Global X U.S. Electrification ETF (WIRE) offers a lower-beta complement to direct data-center exposure, with a beta of just 0.60 versus the S&P 500 and a forward P/E declining from 25.4x in 2025 to 21.7x in 2026 as earnings growth is expected to outpace multiple expansion.
5. 05A suggested portfolio anchors 30% in the WIRE ETF as a diversified core, with Eaton and Vertiv as the largest individual positions (17.5% each) reflecting their critical-bottleneck roles in power distribution and cooling, alongside smaller allocations to Equinix (15%), Arista Networks (10%) and Constellation Energy (10%).

Discussion

The analysis combines institutional infrastructure forecasts from McKinsey, Statista, JLL and the IEA with company-level financial metrics for five companies spanning the data-center value chain, alongside ETF-level fundamentals for the broader electrification theme.

Because data-center buildout and grid modernization are interdependent rather than separate investment cases, the paper treats electrification as the enabling layer beneath AI infrastructure rather than an independent theme, and builds three scenarios around how electricity availability — rather than computing demand — could pace the cycle.