The real estate industry has long been a barometer for economic and technological shifts. The words “digital transformation” say just that: transformation. The global demand for data centres has surged, positioning them as one of the most important asset classes in the current commercial real estate environment. The very nature, importance, and imperative of data centres has changed as technology, political considerations, and technology use has shifted.
Historically, first generation data centres catered to enterprise level IT demands (bespoke corporate facilities, financial services, e-trading, online commerce, etc.). These facilities were, in a post 9/11 world, built for redundancy and hardening of business continuity. Shortly after 9/11, and the “dot com bubble” bursting (remember pets.com?), we recall press stories indicating that “every data centre the world will ever need probably already exists.”
Technology being technology, the world then developed and migrated to cloud computing dispersing the singular criticality of any single facility. Cloud computing seamlessly moved the activity across networks. This technology-based redundancy and risk mitigation reduced the endless pursuit of the “next 9” in facility redundancy and hardness. Facilities became less redundant, somewhat less complicated, and arguably less important and costly in and of itself as technology overcame the physical risks of power loss or even entire loss of a facility. The shift to cloud computing did however expand the footprint of world data centre inventory as eCommerce blossomed, newer blade server technology increased power and cooling density, and the world became ever more interconnected.
In the last five years, the rise of machine learning and artificial intelligence (MLAI) has introduced new challenges and data centre requirements, spawning a new gold rush for appropriate data centre real estate (power and fibre proximity, ample water). As real estate developers, investors, and planners look ahead, they must grapple with evolving energy demands, sustainability concerns, and infrastructure availability to meet industry demand.
The shift from cloud driven demand to third generation data centre MLAI-driven infrastructure has redefined the requirements and contents of data centres. Traditional facilities, which were typically built to 36 to 80 megawatts (MW) of critical capacity, are now being designed for 100 to 200 MW to support the increasing computational power demands of AI workloads. The consumption for a single large language model search reportedly consumes a litre of water for cooling. Ordering a tin of cat food after an eCommerce search does not. This shift has not only changed power consumption patterns but has also raised serious questions about grid stability and resource availability across various global markets.
For real estate developers, this translates into ever more precious developments. Sites capable of meeting requirements for power availability, water, and cooling efficiency have become paramount site selection factors. The axiom ‘location, location, location’ has never been truer, and there simply are not that many suitable locations. Locations with existing power grids capable of handling massive resource demands—or those on site or dedicated power generation potential—are at a premium, leading to increased competition among developers, technology companies, and municipalities worldwide.
Power availability is now one of the most pressing concerns in data centre development. Major data centre hubs, including Singapore, Frankfurt, London, Tokyo, and Sydney, face power shortages that could delay projects by several years. In Europe, the challenge is particularly significant, with some regions requiring up to six years to secure adequate power supply. This timeline mismatch between the commercial imperative to put AI computing capacity online and power infrastructure development lead time forces developers to adopt innovative strategies, such as securing land early, applying for power allocation immediately, and building data centre shells speculatively while awaiting power access. A recent trend has also led clients to repurpose other industrial facilities that already have power to the site.
Additionally, data centres now compete with industries such as manufacturing and electric vehicle production for essential electrical distribution equipment, further exacerbating supply chain bottlenecks. Where diesel power generators once had lead times of 16 to 18 weeks, today, they can require anywhere from 52 to 70 weeks, requiring major producers of data centre space to speculatively procure and store equipment to meet future construction demand. For the real estate sector, this means factoring in not just location but also the massive and early outlay of procurement capital, long-term power agreements, and supplier partnerships.
As the industry faces growing scrutiny and regulation relating to Environmental, Social, and Governance performance, data centres are under growing scrutiny for their immense energy and resource consumption. Governments across regions, particularly in Europe, have begun enacting regulations to curb data centres’ carbon footprints. Germany, for instance, now mandates that new data centres reuse a minimum of 10% of their energy starting in 2026, increasing to 20% by 2028.
While renewable energy sources such as solar and wind are often touted as solutions, their current generation capacities fall far short of what is required to power large-scale data centres. Instead, many developers and operators are pursuing a dual approach: investing in power purchase agreements (PPAs) with renewable energy providers and acquiring adjacent land for conservation purposes to offset environmental impacts.
The race to develop data centres is further complicated by lack of labour in key regions globally, where skilled electrical and mechanical trade workers are in short supply. As a result, the industry is turning to modular construction and prefabrication strategies. Instead of assembling electrical rooms and switchgear on-site, components are now being pre-built, energized, and tested inside controlled facilities before being transported for final connection and commissioning. This approach can accelerate project timelines, making it an attractive option for owners and developers facing project deadlines and escalating costs.
Despite the challenges, data centres remain one of the most exciting asset classes in commercial real estate. If enterprise data centres were “gen one,” cloud computing “gen two,” and the birth of large language model AI “gen three” in the evolution of data centre requirements, is “gen four” already making itself apparent? Much has been in the recent press about reduced computing requirements for the Chinese-built AI platform. But other speculation is that large language model AI is migrating to a reasoning AI model whereby large language model interacts and iterates to refine outcomes among itself, producing outcomes generated by longer and longer string inquiries generated artificially. Rather than a single AI based inquiry querying the LLM, what happens when the query is generative in AI, working across and among multiple LLMs? Rather than take one outcome, the outcome of the query is generatively improved among multiple AI systems. Rather than one liter of water for the query, and a first outcome in less than a minute, what if “reasoning” AI took an hour working among multiple models to improve the outcome? What would the requirement be? Ostensibly future generations of AI data centres could require multiples of the current resource demands on power and cooling.
The race to develop AI is existential for many of the world’s largest technology powers. They will succeed, or arguably become the next “pets.com” or Yahoo! When was your last Yahoo! search? Try a Skype call recently? Real estate developers and investors who proactively address these concerns—through strategic site selection, innovative power solutions, and adaptable construction models—will be best positioned to benefit from this next wave of digital infrastructure expansion.
The future of data centres is about speed to market, technological creativity and imperatives, coupled with a diminishing universe of acceptable real estate, energy resource, and sustainability demands balancing and coming together to redefine what’s possible. Those who recognize this intersection and innovate accordingly will lead the charge in shaping the next generation of digital infrastructure.
Source: itbrief
