Jon Abbott, technologies director - global strategic clients at Vertiv looks at how AI can influence smart buildings.
Artificial Intelligence (AI) is becoming central to how organisations operate. But while attention has focused on what AI can do - from automating customer service to improving diagnostics - a deeper understanding of the infrastructure needed to support it is required.
AI workloads place huge pressure on the physical infrastructure. They draw more power, generate more heat, and behave far less predictably than conventional IT systems. For building operators, facilities managers and infrastructure designers, this shift raises urgent questions.
The conventional model - air-cooled server rooms on standard floor plates - no longer meets the brief. Smart buildings now need to accommodate even more complex infrastructure. This means rethinking how cooling, power and space are designed to keep up with AI’s growing demands.
Cooling requirements are rising fast
At the core of the challenge is heat. AI applications rely on high-performance processors, particularly graphics processing units (GPUs), which produce much more heat than long-established central processing units (CPUs). In addition, for the GPUs to function effectively as a cluster, energy and thermal density increases since they are placed closely together. In some cases, individual racks are reaching densities of 50kW to 150kW, far beyond what standard air-cooled systems were designed for.
This has led to increased interest in liquid cooling - either direct-to-chip or immersion-based systems - as a means to control temperature and reduce energy waste. For smart buildings, this requires a more integrated approach: heating, ventilation and air conditioning (HVAC) systems must work in coordination with IT and power systems, not in isolation.
Facilities with integrated building management systems (BMS) are already better positioned to support this. Those that can adjust airflows, respond to thermal spikes and optimise energy use in real time will be more resilient as AI deployments increase.
Electrical infrastructure needs flexibility
AI doesn’t just change the cooling profile; it also shifts how and when energy is consumed. AI servers often run continuously, with bursts of high-power usage that can exceed normal load expectations. This challenges building electrical systems, especially where power distribution is shared between IT and general services.
To meet these demands, some buildings are adopting software-defined power management systems. These platforms monitor usage patterns, predict peaks and adjust distribution accordingly. They also help manage battery systems and on-site energy generation, which are becoming more common where grid capacity is constrained.
Grid constraints are no longer rare. In some regions, especially near dense urban centres, buildings may face long lead times for increased electrical capacity. Smart infrastructure must account for this, balancing energy use, load scheduling, and backing up systems in real time.
AI is accelerating demand for heat reuse
Ironically, the thermal intensity of AI also creates new opportunities. Some smart buildings are beginning to explore the use of waste heat from IT infrastructure - particularly liquid-cooled systems - for heating other parts of the building or supplying nearby facilities.
This kind of integration has been discussed for some time, but AI’s rising heat output makes it more viable. Where fluid temperature can be maintained at usable levels, and demand is consistent, heat reuse becomes an environmental and economic opportunity.
Smart buildings with heat recovery systems already in place - or the ability to install them - will find themselves ahead of both efficiency targets and upcoming regulations concerning energy performance.
A system-wide rethink
The growing complexity of AI workloads doesn’t just mean more power and cooling. It requires better coordination across building systems. The most successful smart buildings will be those that treat IT infrastructure as part of a broader energy and environmental strategy - not as a sealed room in the basement.
Commissioning processes are also changing. As AI systems vary in usage, infrastructure must be tested under more dynamic conditions. Collaboration between IT, facilities and engineering teams is increasingly critical - not just at handover, but throughout the building’s lifecycle.
As AI becomes embedded in everyday operations, smart buildings will play a vital role in enabling it - quietly, efficiently and intelligently.