“Our experts correctly identified some of the trends that would emerge in 2024, such as the proliferation of AI and the need to transition to more complex liquid and air cooling strategies. In 2025, we expect activity on this front to accelerate and evolve even further,” said Vertiv CEO Giordano (Gio) Albertazzi. “With AI driving rack densities that reach hundreds or even thousands of kW, the need for advanced, scalable solutions to power and cool these racks, minimize their environmental footprint, and empower these emerging AI factories is greater than ever. We anticipate significant progress in this area during 2025, and our customers are demanding it.”
The following are the trends most likely to emerge in the data center industry during 2025, according to Vertiv experts:
1. The power supply and cooling infrastructure will innovate to keep pace with computing densification:
By 2025, the impact of large computing workloads will intensify, and the industry will seek to manage this sudden shift in various ways. Advanced computing will continue its transition from CPUs to GPUs to leverage the latter's parallel computing capabilities and the higher thermal design point of modern microchips. This will continue to put pressure on existing power and cooling systems, forcing data center operators to adopt cold plate and immersion cooling solutions that remove heat at the rack level. Enterprise data centers will be impacted by this trend as AI usage expands beyond cloud and colocation service providers.
AI racks will require uninterruptible power supplies (UPS), batteries, electrical panels, and distribution equipment with higher power densities to handle AI workloads that can fluctuate, in an instant, from 10% (in standby mode) to 150% overloads.
Hybrid cooling systems, with liquid-to-liquid, liquid-to-air, and liquid-to-coolant configurations, will evolve into rack-mounted, edge-mounted, and row-mounted models for deployment in existing applications or new builds.
Increasingly, liquid cooling systems will be combined with their own dedicated, high-density UPS systems to ensure continuous operation.
Servers will be progressively integrated with the necessary infrastructure, including factory-integrated liquid cooling. Ultimately, this will lead to more efficient manufacturing and assembly, faster deployment, smaller equipment footprints, and greater system energy efficiency.
2. Data centers will prioritize energy availability challenges:
The overexpansion of electrical grids and the surge in energy demand are changing how data centers consume electricity. Globally, data centers use an average of 1-2% of the world's electricity, but AI is driving this consumption, and it is likely to rise to 3-4% by 2030. The anticipated increases will place demands on the grid that many providers will be unable to manage, attracting regulatory attention from governments worldwide. This includes potential restrictions on data center construction and energy use, as well as increased costs and carbon emissions that organizations in the sector are striving to control. These pressures are forcing companies to prioritize energy efficiency and sustainability even more than before.
In 2024, we predicted a trend toward alternative energy sources and microgrid deployments, and in 2025 we are seeing an acceleration of this trend, with real movement toward the search for and prioritization of efficient and alternative energy solutions that are new to this segment. Fuel cells and alternative battery technologies are becoming increasingly available as power options for microgrids. In the longer term, several companies are developing small, modular reactors for data centers and other large electricity consumers. These solutions are expected to be available toward the end of the decade. Progress in this field is worth watching closely during 2025.
3. Industry members will collaborate to boost the development of AI factories:
Average rack densities have steadily increased over the past few years, but for an industry that managed an average density of 8.2 kW in 2020, predictions for AI factory racks are around 500 to 1000 kW, or even more in the near future, representing an unprecedented disruption. As a result of these rapid changes, microchip developers, customers, power and cooling infrastructure manufacturers, energy companies, and other industry stakeholders will increase their collaboration to develop and support transparent roadmaps that facilitate AI adoption. This collaboration will extend to AI-driven development tools to accelerate the engineering and production of standardized and tailored designs. In the coming year, microchip manufacturers, infrastructure designers, and customers will further collaborate on manufacturing partnerships that enable true infrastructure and computing integration.
4. AI will make cybersecurity more complex – and also simpler:
The increasing frequency and severity of ransomware attacks has led to new and broader studies of cybersecurity processes and the role of the data center community in preventing such attacks. Last year, one-third of all attacks involved some form of ransomware or extortion, and criminals are leveraging AI tools to scale their operations, extend their reach, and employ more sophisticated approaches. Attacks often begin with the AI-assisted hacking of control systems, embedded devices, or connected hardware and infrastructure systems that were not always designed to meet the same security requirements as other network components. Without proper due diligence, even the most sophisticated data centers can be rendered unusable.
As cybercriminals continue to leverage AI to increase the frequency of their attacks, cybersecurity experts, network administrators, and data center operators will need to keep pace by developing their own sophisticated AI security technologies. While the fundamental principles and best practices of comprehensive protection and extreme due diligence remain the same, the evolving nature, source, and frequency of attacks add new nuances to today's cybersecurity efforts.
5. Industry regulators and governments will address AI applications and energy consumption:
Although our predictions for 2023 focused on government regulation of energy consumption, by 2025 we anticipate the possibility of regulations increasingly addressing the use of this technology. Governments and regulatory bodies worldwide are racing to assess the implications of this technology and develop guidelines for its use. The trend toward sovereign AI—that is, national control or influence over the development, deployment, and regulation of AI, and regulatory frameworks aimed at governing it—are key points in the European Union's Artificial Intelligence Act and China's Cybersecurity Law (CSL) and Secure AI Governance Framework. Denmark recently inaugurated a sovereign AI supercomputer, and many other countries have undertaken their own sovereign AI projects and legislative processes to expand regulatory frameworks, providing a clear indication of the trajectory of this trend. The evolution of some form of regulation is inevitable, and restrictions are highly possible, if not probable.
Initial steps will focus on technology applications, but as the focus on energy and water consumption, as well as greenhouse gas emissions, intensifies, regulation could extend to the types of AI applications and data center resource consumption. By 2025, governance will still be local or regional in scope, rather than global, and the consistency and stringency of its enforcement will vary considerably.
Author: Mario Vasconcelos, Sales Director Enterprise Accounts, Spain & Portugal at Vertiv
