Data Center Demand Outpaces Supply Despite Efficiency Gains

Written by:

Steve Marcinuk

Date:

08 Sep 2025

Share

Share via Facebook

Share via Linkedin

Share via X

Copy to clipboard

The data center sector continues to experience rapid growth driven by artificial intelligence applications, despite recent concerns about potential overbuilding and computational efficiency breakthroughs. Industry analysts suggest that current supply additions, while substantial, are largely pre-committed and insufficient to meet emerging demand patterns.

The current U.S. data center landscape encompasses approximately 30 gigawatts of capacity, with projections indicating growth to 50 gigawatts over the next five years. While construction activity appears robust with eight to nine gigawatts currently under development, much of this capacity is being built by hyperscalers for their own operations.

“A large chunk of that is actually being built by owner users or owner operators, the hyperscalers themselves. And if they’re building it themselves for their own purposes, it’s almost pre-leased,” explains Vikram Malhotra, Managing Director and Co-Head at Mizuho Americas, who covers data centers among other real estate sectors.

Early Stages of AI Evolution

The artificial intelligence growth driving data center demand remains in its early stages, according to industry observers. Current AI applications like ChatGPT and Perplexity represent basic search and analytics functions rather than true machine learning capabilities.

“I personally view it as Google Search and Google Analytics combined,” Malhotra notes. “If I ask ChatGPT to project warehouse demand for 2026, you can see what it’s searching, essentially trolling the internet, looking at Jones Lang LaSalle, CBRE, Prologis for that answer, rather than using data sets to come up with a number.”

This early-stage development suggests significant room for growth as AI capabilities advance. The evolution from current applications to more sophisticated predictive models could drive substantial additional demand for computing infrastructure.

Private AI Models Drive Distributed Demand

Beyond public AI applications, the emergence of private artificial intelligence models presents another growth driver. Large corporations are expected to develop proprietary AI systems tailored to specific business needs.

“I would imagine a Walmart, a big bank, a big manufacturer would want their own private model for their own business purposes, and not just have some generic models,” Malhotra explains. This trend toward private AI could drive demand for both large-scale data centers and smaller, distributed facilities.

The distinction between large language models (LLMs) and smaller language models (SLMs) creates diverse infrastructure requirements, potentially supporting various data center configurations and locations.

Power Constraints Create Supply Bottlenecks

Despite construction activity, power availability has emerged as a critical constraint limiting new data center development. The infrastructure required to support high-density computing creates significant barriers for new market entrants.

“If a data center or landlord does not have power contracted, and you’re a new player coming in, best of luck getting power before 2028 or 2029,” Malhotra observes. “There’s a real demand-supply imbalance today, and I think it’s going to start to translate into pricing power for the landlords.”

This power constraint effectively limits competition and supports pricing strength for established operators with secured power arrangements, mirroring traditional real estate dynamics where supply constraints drive rental growth.

Efficiency Gains Provide Temporary Headwinds

Recent breakthroughs in computational efficiency from international AI companies have raised questions about future infrastructure requirements. These advances show that similar AI capabilities can be achieved with reduced computing resources.

However, industry analysts view efficiency gains as temporary rather than fundamental demand reducers. The analogy to smartphone evolution suggests that while individual applications become more efficient, overall usage and capability expansion drive continued infrastructure demand.

“What I’m talking about is the next new foundational model that becomes even smarter, with even more self-predictability,” Malhotra explains. The development of more sophisticated AI systems is expected to offset efficiency gains through expanded capabilities.

Hyperscaler Spending Patterns

The sustainability of current growth depends heavily on continued capital expenditure from major technology companies. Hyperscalers currently deploy hundreds of billions annually in infrastructure investments, with data centers representing a significant portion.

Market participants closely monitor hyperscaler spending announcements for signals about future demand. Historical precedent from the warehouse sector, where Amazon’s 2022 announcement of footprint rationalization preceded sector underperformance, illustrates the impact of major user spending changes.

“If the Googles, the Metas, and Amazons come out and say we’ve built out our data center network and we’re now going to rationalize, that may be a sign that things are overbuilt in the very near term,” Malhotra notes.

Investment Implications and Market Dynamics

The data center sector presents compelling investment characteristics despite near-term volatility concerns. Companies like Digital Realty Trust are positioned to benefit from the transition from minimal earnings growth to consistent high single-digit growth over the next two to three years.

Key monitoring factors include hyperscaler capital expenditure outlooks, enterprise cloud adoption rates, and inference demand growth. The evolution toward edge computing and distributed AI processing could create additional demand categories beyond traditional centralized data centers.

Pricing power emergence is a critical inflection point for the sector. As vacancy rates approach low single digits in key markets, rental growth acceleration becomes increasingly likely, following traditional real estate market dynamics.

Looking Forward

The data center sector appears positioned for sustained growth despite periodic concerns about overbuilding or efficiency improvements. The combination of early-stage AI development, private model adoption, and power supply constraints creates a favorable supply-demand environment.

While fluctuations are expected as technology evolves and spending patterns shift, the fundamental drivers supporting data center demand remain intact. The sector’s evolution from a niche infrastructure play to a critical component of the digital economy continues to unfold, with significant implications for real estate investors and technology companies alike.

The intersection of artificial intelligence advancement, corporate digital transformation, and infrastructure constraints positions data centers as a key beneficiary of ongoing technological change, despite periodic market concerns about pace and sustainability of growth.