Challenges for data centers in 2025: VERTIV indicates the commitments of the sector for the support, activation, development and regulation of AI
Innovation in AI rack power and cooling systems, energy consumption and emissions management are in focus in the new year
Artificial intelligence continues to transform the data center industry, a reality reflected in the trends predicted for 2025 by Vertiv, (NYSE: VRT), a global provider of critical digital infrastructure and continuity solutions. Vertiv experts predict an increase in the industry's level of innovation and integration to support high-density computing, regulatory scrutiny of AI, and a growing focus on sustainability and cybersecurity initiatives.
“Our experts have correctly identified the proliferation of AI and the need to adopt more complex liquid and air cooling strategies as one of the trends in 2024. Activity on this front is expected to further accelerate and develop in 2025,” said Giordano Albertazzi, CEO of Vertiv. "With AI increasing rack densities by three or four digits of kW, the need for advanced, scalable solutions to power and cool these racks, minimize their environmental impact, and enable emerging AI Factories has never been greater. We expect significant progress on this front in 2025, as requested by our customers."
Vertiv experts indicate the trends that could emerge in data centers in 2025:
- Power and cooling infrastructure is revamped to keep pace with computing densification. Advanced computing will continue to move from the CPU to the GPU to take advantage of the latter's parallel computing power and the high level of thermal design of modern chips. This will further challenge existing power and cooling systems and push data center operators towards cold-plate and immersion cooling solutions, which remove heat at the rack level. Enterprise data centers will be affected by this trend, as the use of AI extends beyond the primary cloud and colocation providers.
- AI racks require uninterruptible power supplies, batteries, power distribution devices and switchgear with higher power density to handle AI loads that can go from 10% down to 150% overload in a flash.
- Hybrid cooling systems, with liquid-liquid, liquid-air and liquid-refrigerant configurations, will evolve into rack, edge and array cabinet designs that can be used in brown/greenfield applications.
- Liquid cooling systems will increasingly be combined with dedicated high-density uninterruptible power supplies to ensure continuous operation.
- The servers will be integrated into the infrastructure that supports them, including the liquid cooling system integrated into the design phase, making manufacturing and assembly more efficient, deployment faster, equipment footprint smaller and increasing energy system efficiency.
- Data centers prioritize energy availability challenges. Overloaded networks and increasing energy demand are changing the way data centers use energy. Globally, data centers consume an average of 1-2% of the world's electricity, but AI is driving an increase in demand that will likely reach 3-4% by 2030. The expected increases could result in energy demands on the grid that many utilities cannot handle, drawing the attention of governments around the world – including potential restrictions on data center construction and energy use – and driving up costs and CO₂ emissions that data center companies are trying to contain. These pressures force companies to prioritize energy efficiency and sustainability much more than in the past.
In 2024 we predicted a growth in alternative energies and microgrids and in 2025 we are seeing an acceleration of this trend, with a real movement towards the priority and search for energy efficient solutions and new energy alternatives in this field. Fuel cells and battery alternatives are increasingly available as energy options for microgrids. In perspective, many companies are developing small modular reactors for data centers and other large energy users, expected to be available by the end of the decade. Progress on this front is to be observed in 2025.
- Industry operators collaborate to encourage the development of the AI Factory. Average rack density has been steadily increasing in recent years, but for an industry that supported an average density of 8.2kW in 2020, predictions of an AI Factory rack of 500 to 1000kW or more represent an absolute innovation. As a result of rapid evolution, chipmakers, customers, power and cooling infrastructure providers, utilities and other industry stakeholders will increasingly interact to develop and support robust roadmaps to enable AI adoption. The collaboration extends to AI-powered development tools to accelerate the design and manufacturing of standardized and customized components. In the coming year, chipmakers, infrastructure designers and their customers will increasingly collaborate and move towards industry partnerships that enable true integration between IT and infrastructure.
- Artificial intelligence makes cybersecurity more complex – and simpler at the same time. The increase in frequency and severity of ransomware attacks is driving a new and broader consideration of cybersecurity processes and the contribution that the data center community can make to preventing such events. A third of all attacks last year involved some form of ransomware or extortion, and today bad actors are leveraging AI tools to intensify assaults, broaden reach and employ sophisticated approaches. Attacks increasingly begin with AI-supported hacking of control systems, embedded devices or connected hardware and infrastructure, which are not always designed to meet the same security requirements as other network components. Without adequate attention, even the most advanced data center can become unusable.
As cybercriminals continue to leverage AI to intensify the frequency of attacks, cybersecurity experts, network administrators and data center operators must keep pace by developing their most advanced AI security technologies. Although the cornerstones and best practices of integrated defense and extreme accuracy remain unchanged, the change in terms of the type, origin and frequency of attacks adds complexity to cybersecurity initiatives.
- Government and industry regulations address the application of AI and energy use. While our predictions for 2023 focused on government regulations for energy use, by 2025 we expect regulations to increasingly address the use of AI itself. Governments and regulators around the world are accelerating the assessment of the implications of AI and the development of governance for its use. The move towards sovereign AI – a nation's control or influence over the development, deployment and regulation of AI, and regulatory frameworks aimed at governing AI – is at the heart of the European Union's Artificial Intelligence Law and China's Cybersecurity Law (CSL) and AI Security Governance Framework. There Denmark recently inaugurated its own sovereign AI supercomputer and many other countries have undertaken sovereign AI projects and legislative processes to establish regulatory frameworks, an indication of the trajectory of this trend. Some form of regulation is inevitable and restrictions are possible, if not probable.
The initial phases will focus on the application of the technology, but as the focus on energy and water consumption and greenhouse gas emissions intensifies, the regulations could also extend to AI applications and data center resource consumption. In 2025, governance will continue to be local or regional rather than global, and the consistency and rigor of enforcement will vary widely.
