Exponential demand on AI computing capacity has triggered a boom in the data center market: current facilities are going through expansive refactoring, but there is also unparalleled demand to build new facilities. This trend is happening globally with the U.S., China and Europe re-evaluating their computing footprint to gain a leadership position in the race for AI domination. This has created ripple effects on the entire data center supply chain, real estate being only one aspect.
On one hand, hyperscalers apply their dominant power to secure the best possible sites with potential ownership over the entire value chain, including power generation. On the other hand, second-tier sites also benefit from the demand. This comes with a necessary reevaluation of real estate assets, how they are priced and financed in the context of a technology revolution and economic power battleground.
Data centers as strategic infrastructure
A couple of decades ago, data centers and major communication hubs would have looked like warehouses, hosting rows and floors of servers, not unlike distribution warehouses and logistics hubs. Square footage, occupancy rates and power capacity were some of the key metrics.
With AI workloads becoming more demanding for every new model announcement, additional metrics must be considered. Several of those are, unfortunately, showing the emergence of bottlenecks in the race towards ubiquitous AI. When looking at recent announcements, the value of data center facilities is now ranked based on:
- Strategic access to the energy grid or electricity production sites
- Proximity to fiber and advanced communication hubs
- Availability of GPUs and GPU acceleration techniques
- Energy efficiency and cooling systems
- Ability to comply with domestic and international regulatory requirements
Compute, network, or storage demands from AI workloads such as large language models (LLMs), model training, inference workloads and data preparation processes can require significantly more energy than traditional computing requests. For example, AI Deep Search compute and energy costs are an order of magnitude higher than for traditional Google Search queries.
Energy dominates the data center capacity discussion
Data center facilities don’t just need power, they need it to be stable, reliable, and resilient to physical events and cyber attacks. As of now, this has become the primary factor in evaluating data center facilities. Traditionally, in the U.S.,facilities have been concentrated in:
- Virginia: close to D.C., major communication hubs that have hosted telecommunication providers for decades.
- Silicon Valley: central technology hub but potentially at risk of natural disaster (earthquakes).
- Texas: central U.S. location, also a hub for telecommunication, affordable land and energy, with additional tax incentives.
It is worth noting that Europe, because of its high density, faces more constraints regarding expanding data centers, especially given the location of traditional sites next to major cities like London, Frankfurt, or Amsterdam.
Location constraints are one of the main reasons why infrastructure investors are making moves towards second-tier locations that either remove power limitations or provide an economically viable path for the build-out of the entire supply chain necessary to support modern data centers.
GPUs: Accessibility and readiness
Traditional data centers are built around CPUs. AI workloads depend on Graphical Processing Units (GPUs), whose production is dominated by NVIDIA, and Tensor Processing Units (TPUs), which are proprietary to Google. These processing units are expensive, backlogged due to limited fabrication capacity, and even gated based on countries’ strategic interests.
In parallel, deploying GPUs in a concentrated environment results in potential meltdown (we have read about OpenAI challenges after the release of its AI image generation in March 2025). New generation cooling technologies based on liquid cooling are necessary for dense AI data center facilities and are one of the reasons why it is not always feasible to retrofit existing facilities. A ready-for-use facility equipped with state-of-the-art cooling mechanisms, high-density CPUs or TPUs commands a premium from tenants and investors.
The cloud has centralized computing, but users have not moved
Cloud computing drastically simplified system and application deployments, starting with the geographic distribution of services and data, but with one constraint: the network needed to support the load. It took decades for the network to evolve from 2G to 5G and continue to support the load of ever more sophisticated applications. Will AI computing follow a similar journey?
AI workloads with direct user interaction (chatbots, real-time translation), inference models deployed on edge systems, or chained models, for example, are more sensitive to latency than others. Fiber connectivity to and between AI hubs will emerge as a severe bottleneck, adding another factor to evaluating AI-ready data center facilities. Communications will be a challenge, and cybersecurity issues will be right behind.
From regulatory pressures to sovereignty battles
The ability to transport large volumes of data at speed, without compromising security and compliance requirements, is foundational and influences the selection of sites and the ability to attract tenants.
The location of data center facilities has become essential to national AI strategies, data sovereignty and the resilience of most countries’ digital economies. Many entities and governments with the capacity to influence, control, or invest in data center facilities are realigning their priorities to secure a path for immediate and future AI initiatives:
- Governments are increasing their interest in AI infrastructure, encouraging and supporting the private sector to pursue AI-focused real estate investments. (e.g., The U.S. Department of Energy (DOE) has recently released a Request for Information (RFI) to inform possible use of DOE land for artificial intelligence (AI) infrastructure development.
- Hyperscalers like AWS, Google and Microsoft that already operate a large footprint of facilities on a global scale are rapidly seeking full ownership of the entire data center supply chain through investments and acquisitions.
- Energy companies have proactively developed energy alternatives for data centers through renewable energy investments.
- Real Estate Investment Trusts (REITS) and infrastructure funds are developing new types of funds specially adapted to AI-ready data center assets.
Bottom line: The strategic value of data centers
Nothing is constant in this new era, given the rapid pace at which AI evolves. The entire digital economy is being up-leveled to a new model where access to AI-ready data center facilities becomes the prize. Today, dollars per megawatt might be the most common currency to price data center facilities. With the increasing strategic importance of AI-ready facilities, all stakeholders need to be on the watch to extend traditional real estate metrics to those that reflect the elevated nature of such facilities. These include — power, GPUs, new design, communication and sovereignty.
