AI data center capacity has grown 50-fold in four years, pushing infrastructure ETFs like PAVE to the forefront of the build-out trade.
AI data center capacity has grown 50-fold in four years, pushing infrastructure ETFs like PAVE to the forefront of the build-out trade.

AI data center capacity has grown 50-fold in four years, pushing infrastructure ETFs like PAVE to the forefront of the build-out trade.
AI data center capacity has expanded 50 times in four years, driving demand for grid equipment and power infrastructure that is lifting the Global X U.S. Infrastructure Development ETF (PAVE) as utilities capture rising build-out spending.
"The first phase of the AI trade was about chips. The next phase continues to be about the build," Jay Woods, chief market technician at Chase Games, said in a CNBC Pro analysis.
MasTec Inc., a builder of transmission lines and electrical grids, has surged 133% over the past 52 weeks as utilities and grid equipment firms benefit from rising AI-related capital expenditure. Data center sizes have grown 50-fold in four years, according to Exponential View. Token costs have fallen 60% to 70% annually by Goldman Sachs' estimate, yet nearly three-quarters of enterprises exceeded their AI cost projections last year, the FinOps Foundation found.
The infrastructure build represents a multiyear spending cycle extending beyond mega-cap technology names. PAVE offers diversified exposure to utilities, grid equipment, and construction firms positioned to benefit as hyperscalers pour capital into power-hungry data centers. Goldman Sachs projects token consumption will grow 24 times by 2030, requiring commensurate expansion in physical infrastructure.
Grid Spending Becomes the Next AI Bottleneck
The rapid expansion of artificial intelligence has created a secondary investment theme: the physical infrastructure required to power and connect data centers. While Nvidia Corp. and other chipmakers captured the first wave of AI spending, the build-out of electrical grids, transmission lines, and backup power systems is emerging as the next constraint.
MasTec, which constructs transmission lines and fiber networks, has seen its shares climb 133% over the past year. The stock recently recaptured its 50-day moving average at $385, with its moving average convergence-divergence indicator triggering a buy signal that last preceded a leg higher, according to Woods. Caterpillar Inc. and Generac Holdings Inc. have also benefited from the infrastructure theme, with Caterpillar's record backlog reflecting demand for heavy machinery used in data center construction.
Storage Infrastructure Emerges as a New Frontier
Beyond power and grid equipment, AI infrastructure spending is expanding into storage purpose-built for inference workloads. Dell Technologies Inc. has demonstrated that offloading KV cache from GPU memory to purpose-built storage can reduce time-to-first-token from 17 seconds to one second at a 131,000-token context window, according to benchmarks on four Nvidia H100 GPUs. The improvement addresses a growing bottleneck as agentic AI models maintain context across extended sessions, consuming gigabytes of GPU memory that could otherwise be used for computation.
Investment Implications
For investors, the infrastructure theme offers exposure beyond the Magnificent Seven. PAVE holds positions in utilities, construction, and industrial firms that are less correlated with semiconductor cycles. MasTec trades with a market capitalization of roughly $20 billion, while Caterpillar and Generac provide additional diversification across the build-out value chain. The key risk is execution: data center delays could push out revenue recognition for infrastructure contractors, though the secular demand trend remains intact.
This article is for informational purposes only and does not constitute investment advice.