Yonggang Wen, Ph.D.
As we move into 2025, the data center (DC) industry is believed to stand at the crossroads of technological innovations, environmental responsibility, and operational transformation. With the increasing demand for high-performance computing resulting from the surge in AI workloads, data centers are no longer just infrastructure providers; they have become global mission-critical enablers of digital transformation and innovation. In this article, building on our decade-long research experience in data center research and development, we share our views of the TOP 10 trends shaping the future of data centers, offering insights into the innovations and challenges driving this critical sector forward.
1. Sustainability and carbon neutrality as the prerequisite of new/retrofit data centers
The transition to carbon neutrality has become a fundamental requirement for the data center industry. Governments, enterprises, and customers are placing unprecedented pressure on data centers to minimize their environmental footprint. This trend has accelerated the adoption of renewable energy sources, such as solar, wind, and hydroelectric power, to meet energy demands sustainably. In the EU, the Climate Neutral Data Center (CNDC) pact is a pledge of industry players and trade associations of cloud infrastructure services and data centers in Europe to achieve carbon neutrality by 2030. One key focus of Singapore’s latest Green Mark for Data Center (GMDC: 2024) is energy efficiency in data center operations and sustainability aspects including intelligent systems, carbon reduction, health and well-being, resilience, and maintainability.
Retrofit projects, with a total capacity shadowing the most optimistic estimation of green built, are equally critical, focusing on energy-efficient upgrades for legacy infrastructure. Operators are integrating advanced technologies like heat reuse systems and AI-native energy management platforms to achieve sustainability goals. Moreover, more and more operators are tracking and reporting detailed sustainability metrics (e.g., PUE, WUE, and CUE, etc), aiming to achieve green certifications such as LEED, BREEAM, and GMDC in Singapore. In 2025, data centers are not only expected to be carbon-neutral but also to contribute positively to their surrounding environments through circular economy practices and partnerships in local energy grids.
2. Rise of AIDC with evolving training/inference dynamic
AI-driven data centers (AIDC) are emerging as a cornerstone of the digital economy. Traditionally, the training phase of AI models has been more resource-intensive than the inference phase. However, this dynamic is reversing as large-scale inference applications, particularly for applications like recommendation systems and generative AI, surpass training in computational and energy demands. It is expected that the split between training and inference workloads will shift from 80:20 in 2023 to 20:80 in 2028, with an equal distribution in 2025.
To meet these challenges, data centers are deploying specialized hardware such as GPUs, TPUs, and custom AI accelerators designed for energy-efficient training and inference. AI orchestration tools are also being employed to dynamically allocate resources between training and inference workloads, ensuring optimal efficiency. In this regard, we predict that more deployment will leverage out-of-band data processing units (e.g., DPU) to regulate compute-intensive GPUs to optimize training and inference workloads. The AIDC of 2025 will prioritize modular architectures that allow seamless scaling to meet the fluctuating demands of AI workloads.
3. Booming of new cooling technologies for AIDC
The heat generated by high-density AI workloads has driven the rapid adoption of innovative cooling technologies. Specifically, we have seen that the power density per rack has increased by more than 10 times, from 3~5KW per rack in IDC to 40-70KW per rack in AIDC. This challenge in heat removal will be further exacerbated by the booming data center capacities built in ASEAN and MENA regions, where the weather conditions are not ideal to data center operations. As a result, liquid cooling and immersion cooling systems are no longer experimental—they are becoming industry standards. These methods provide superior heat dissipation compared to traditional air cooling, enabling data centers to maintain optimal performance while reducing energy costs.
In 2025, we will see further advancements in hybrid cooling systems that combine liquid and air cooling for maximum efficiency and low total cost of ownership (TCO). Additionally, AI-powered cooling systems will gain traction, using real-time analytics and emerging technologies, such as AI4CFD and digital twin, to predict and optimize cooling needs dynamically. Such innovations are critical to ensuring the sustainability and scalability of AIDC operations.
4. Large Language Models (LLMs) emerging as DC Operating System (OS)
Large Language Models (LLMs), in the form of Agents, are evolving to become foundational components for data center operations and management. With their ability to process vast amounts of data and generate actionable insights, LLMs are being leveraged as virtual operating systems for data centers.
These models are being integrated into workflows to optimize energy use, manage workloads, and enhance cybersecurity. By 2025, LLMs will enable autonomous data center management, allowing for predictive maintenance, intelligent load balancing, and real-time anomaly detection. One pilot application we expect to hit the market in 2025 would be the incident response, assisting the operational staff in troubleshooting, guiding the operators with specific actions based on the massive operational knowledge it learns, and even automating the incident response. Our group has also released some AI agents in the OpenAI store, specifically for the Singapore Green Data Green Mark advisor. This shift will significantly reduce operational complexity while enhancing overall efficiency. More importantly, LLMs are the natural choice to mitigate the high turnover of DC operating staff, as experienced by the industry in light of a huge sectorial growth lately.
5. Fusion of IT and OT
The convergence of Information Technology (IT: compute, network, and storage) and Operational Technology (OT: Mechanical & Electrical Systems) is revolutionizing how data centers are managed. Traditionally siloed, these domains are now being integrated to create unified, intelligent systems capable of monitoring and controlling all aspects of data center operations. In 2025, this fusion will enable real-time data sharing between IT and OT systems, allowing for enhanced decision-making and improved resilience. For example, digital twins, powered by this integration, will provide a holistic view of the data center, enabling proactive management and rapid response to operational challenges. Though the slow rhythm of IT-OT fusion is mainly because of business models and/or corporate organizations in co-locators and hyperscalers, in 2025 we expect to see new models in particular for AIDC and AI factories, for the various stakeholders to collaborate.
The integration of IT and OT will also enhance energy management within data centers, enabling dynamic adjustments based on workload demands and environmental conditions. This capability will contribute to improved energy efficiency and sustainability, aligning with global carbon neutrality goals. By bridging these two domains, data centers will achieve greater operational synergy and adaptability, ensuring robust performance in an era of increasing complexity.
Furthermore, IT-OT convergence will drive innovation in predictive analytics and automation. With unified systems, data centers can leverage machine learning models to anticipate potential issues and implement corrective actions in real time. This integration approach is particularly beneficial to the AIDC operations, where the headroom for resilience is reduced to the minimum due to extremely high rack density and low peak-to-average utilization ratio. This proactive approach reduces downtime and enhances reliability, solidifying the role of IT-OT fusion as a cornerstone of modern data center management.
6. Cognitive digital twin for advanced DC operations and management
Digital twin, originally designed for a multi-scale multi-physics simulation capability, has matured into a cognitive tool capable of learning, predicting, and optimizing data center operations. These virtual replicas of physical systems are being enhanced with AI capabilities, transforming them into cognitive digital twins. Following this trend, at NTU and Red Dot Analytics (RDA), we have developed a disruptive World Fusion Model, which presents a cyber-physics continuum to model data centers, for descriptive, predictive, and prescriptive analytic capabilities, called Fusion Intelligence.
By 2025, cognitive digital twins will enable (near)-real-time simulation and predictive maintenance, reducing downtime and operational costs for mission-critical infrastructure, such as IDC and AIDC. They will also play a critical role in sustainability efforts by modeling energy usage and suggesting optimizations to minimize environmental impact. This evolution marks a significant leap in how data centers are designed, operated, and maintained.
7. Green computing in data centers: energy-efficient computing and green power
Green computing is reshaping data centers by emphasizing energy-efficient technologies and the use of clean energy sources. Advanced processors and networking/storage systems that consume less power are being adopted widely, reducing the energy demands of high-performance computing tasks. Simultaneously, data centers are transitioning to clean energy solutions, including solar, wind, hydropower and etc, to align with sustainability objectives.
Beyond hardware, 2025 will see a significant push toward software and algorithm optimization for green computing. Software-defined green computing approach will start to find its deployment in IDC, AIDC and even crypto-mining data centers (CDC). AI-driven algorithms are being streamlined to deliver maximum computational output with minimal energy consumption. This holistic approach—combining energy-efficient infrastructure with intelligent software design—ensures that data centers achieve operational excellence while supporting global environmental goals. The integration of smart energy management systems will further enhance resource efficiency, making green computing a defining feature of modern data centers.
8. Data center as a catalyst for end-to-end energy systems
Data centers are evolving beyond their traditional roles to become active participants in end-to-end energy systems. By integrating with local grids and renewable energy sources, they are enabling smarter energy management across entire ecosystems. Data centers in 2025 will serve as energy hubs, actively balancing supply and demand. Through advanced energy storage solutions such as grid-scale batteries, they will store surplus energy during off-peak hours and release it during peak periods, thereby stabilizing the grid. The data center also presents a golden opportunity for demand response in grid stabilization. These capabilities are particularly valuable in regions with a high penetration of renewable energy, where supply can be intermittent.
In the meantime, AI-native energy orchestration will further enhance this role, optimizing the allocation of energy resources across multiple stakeholders. By analyzing real-time data and leveraging the digital twin technology, AI can predict energy demand and dynamically adjust the flow of electricity to ensure efficiency and cost-effectiveness. This approach not only improves grid reliability but also minimizes waste.
Moreover, data centers are expected to foster collaborations with utility companies and other industries to create integrated energy ecosystems. These partnerships will enable the sharing of energy resources, such as heat generated by servers being repurposed for district heating systems. This holistic approach positions data centers as central players in the transition to sustainable and resilient energy systems.
9. Quantum computing integration
Quantum computing is poised to become an integral part of data center operations in 2025. As the technology transitions from research labs to practical applications, hybrid data centers combining classical and quantum computing will emerge. These centers will excel in solving complex problems in optimization, cryptography, and scientific research, areas where quantum computing offers a significant advantage.
To support this integration, data centers will adopt specialized infrastructure, including cryogenic cooling systems and quantum-safe encryption to secure sensitive data. Additionally, advancements in quantum algorithms and error correction will enhance the reliability and scalability of quantum systems. By incorporating quantum computing, data centers will push the boundaries of innovation, unlocking new capabilities for industries ranging from healthcare to logistics.
10. Interoperability and data sovereignty
With increasing global regulations around data sovereignty, ensuring compliance while maintaining interoperability will be a key challenge for data centers in 2025. Operators will need to navigate complex regulatory landscapes while enabling seamless data exchange across borders.
Challenges always present opportunities for growth and development. This will drive the adoption of standardized frameworks and secure data-sharing protocols. AI and blockchain technologies will also play a role in ensuring compliance and transparency. By prioritizing interoperability and sovereignty, data centers will build trust and maintain their competitive edge in an increasingly interconnected world.
As these trends demonstrate, 2025 promises to be a transformative year for the data center industry. From sustainability to quantum computing, the innovations shaping this industry will not only address current challenges but also open new frontiers for technological advancement. By embracing these trends, data centers will continue to drive progress and empower the digital economy.
About the author
Dr. Yonggang Wen is a Professor and President’s Chair in Computer Science and Engineering at Nanyang Technological University (NTU), Singapore. He serves as the Associate Provost (Graduate Education) and Dean of Graduate College at NTU Singapore. He is also the Chairman and Chief Scientist at Red Dot Analytics Pte Ltd, an NTU spin-off, focusing on AI and digital infrastructure. He received his PhD degree in Electrical Engineering and Computer Science (minor in Western Literature) from Massachusetts Institute of Technology (MIT), Cambridge, USA, in 2008. Dr. Wen has published over 300 papers in top journals and prestigious conferences. His systems research has gained global recognition. His work in Multi-Screen Cloud Social TV has been featured by global media (more than 1600 news articles from over 29 countries) and received the ASEAN ICT Award 2013 (Gold Medal). His work on Cognitive Digital Twin for Data Centre, has won the 2015 Data Centre Dynamics Awards – APAC (the ‘Oscar’ award of the data centre industry), the 2016 ASEAN ICT Awards (Gold Medal), the 2020 IEEE TCCPS Industrial Technical Excellence Award, the 2021 W.Media APAC Cloud and Datacenter Technology Leader Award, and 2022 Singapore Computer Society Digital Achiever Tech Leader Award. He was the winner of the 2019 Nanyang Research Award and the sole winner of the 2016 Nanyang Awards for Innovation and Entrepreneurship, both of which are the highest recognition at NTU. He is a co-recipient of multiple Best Paper Awards from top journals, including 2019 IEEE TCSVT and 2015 IEEE Multimedia, and at international conferences, including 2023 ASPLOS, 2016 IEEE Globecom, 2016 IEEE Infocom MuSIC Workshop, 2015 EAI Chinacom, 2014 IEEE WCSP, 2013 IEEE Globecom and 2012 IEEE EUC. Currently, he is the Editor in Chief of IEEE Transactions on Multimedia (TMM), serves or has served on editorial boards for multiple IEEE and ACM transactions, and was the elected Chair of IEEE ComSoc Multimedia Communication Technical Committee (2014-2016). His research interests include cloud computing, green data centers, big data analytics, multimedia networks, and mobile computing. He is a Fellow of IEEE and the Singapore Academy of Engineering.