Artificial intelligence is becoming embedded in healthcare operations through tools such as ambient listening, productivity features and diagnostic algorithms. As these capabilities expand, they increase demand for computing capacity and place new pressure on power and cooling infrastructure. Greater computing requirements also add to strain on the energy grid, alongside the impact of more severe weather events, and rising demand contributes to higher energy costs. At the same time, healthcare data centres must protect increasingly valuable hardware with cooling approaches that can manage denser loads. Power continuity and thermal management therefore become practical constraints on how quickly AI-enabled workflows can scale across sites and services.
Strengthening Power Resilience for AI Workloads
AI workloads increase the importance of extended runtime for critical systems, particularly during disruption. Healthcare organisations may rely on generators, but regulatory requirements can also mean that battery energy storage systems are necessary. Charging a battery energy storage system overnight at lower electricity rates and drawing on it during the day allows the building’s supply to be supplemented when demand and costs are higher. Whether an organisation uses a generator, a battery system, an uninterruptible power supply (UPS) or a combination, the core requirement is enough runtime to support the power load associated with AI and other clinical applications.
Must Read: Securing AI Agents Against System-Level Threats
Power considerations also vary by location within the care environment. Where UPS equipment is installed in patient care spaces, medical-grade systems are needed to avoid emission of specific electromagnetic waves and to maintain safety around patients. These requirements shape both equipment selection and where power protection can be deployed. Planning also needs to extend beyond individual devices to facility-level prioritisation. Health systems require a clear strategy identifying which facilities and services will continue to receive power during natural disasters or other crises that interrupt primary energy sources. That approach supports continuity for essential operations that depend on digital systems, including AI-enabled processes.
Reworking Cooling for Denser Data Centre Loads
As AI drives higher computing density, it increases heat generation and can exceed what traditional air-based cooling can manage on its own. Conventional data centres depend on effective air flow and rapid management of hot spots to protect equipment, but denser AI workloads raise the threshold for thermal control. In response, organisations are beginning to add liquid cooling to address higher heat loads. Current deployments often use hybrid designs, with liquid cooling applied directly to processors and sometimes memory, while other components continue to be cooled by fans. The trajectory of AI adoption points towards a shift from hybrid environments to fully liquid-cooled settings.
Cooling changes are closely linked to wider infrastructure decisions. Introducing liquid cooling requires planning for how it will be integrated and where it will be hosted, including decisions about retrofitting existing data centres or using prefabricated facilities or colocation services. Colocation is widely used, but it can be costly, and a return towards on-premises data centres is anticipated. Prefabricated data centres provide another pathway because they can be located outside existing facilities, avoiding complex retrofits. Converting an indoor data centre to a prefabricated model can also free space within hospitals for additional beds, which remain a key source of revenue for healthcare organisations.
Sustainability and Safety in Energy and Cooling Choices
Power and cooling choices intersect with sustainability goals, equipment lifecycle planning and safety considerations. Battery technology illustrates these trade-offs. Traditional lead acid batteries typically last three to five years. Lithium-ion batteries can operate for seven to ten years, improving environmental performance and total cost of ownership. They do not require maintenance and, when replacement is due, it is likely to coincide with a market in which more advanced options are available. Even so, adoption of lithium-ion batteries in healthcare has been slower due to concerns influenced by widely shared images of electronic devices catching fire. Organisations can be cautious about deploying these batteries in environments where flammable oxygen is present. New battery chemistries with fewer fire-related concerns are expected to make the advantages of longer life and reduced maintenance more prominent.
Liquid cooling also raises environmental questions that need to be understood alongside performance requirements. There are four main approaches: direct liquid cooling and immersion cooling, each available in single-phase and two-phase configurations. Single-phase systems typically use water or water mixtures, while two-phase systems use refrigerants that change to gas when heated. Across these approaches, organisations need clarity on the environmental impact of water, chemicals, heat or vapours discharged back into water systems or the atmosphere. Establishing how each option behaves in real-world deployment supports decisions that balance capacity, resilience, safety and environmental impact as AI workloads expand.
AI adoption is reshaping infrastructure expectations in healthcare by increasing the importance of power continuity and more capable cooling for data centres. Extended runtime, including the use of battery energy storage systems and appropriate UPS configurations, supports operational resilience when primary energy sources are disrupted. Denser computing loads are also accelerating a move from air-only cooling towards hybrid and potentially fully liquid-cooled environments, with implications for whether organisations retrofit, use prefabricated facilities or rely on colocation. Battery lifecycles, safety concerns and the environmental effects of liquid cooling approaches remain central considerations as health systems plan for sustained AI-enabled growth.
Source: HealthTech
Image Credit: iStock
