AI’s Growing Dependence on Data Centers: Power and Cooling Solutions Must Evolve
The rapid growth of artificial intelligence (AI) and its various applications, including generative AI (GenAI), has highlighted the increasing importance of data centers and edge networks. These infrastructures play a crucial role in enabling the magic that emerges from AI applications by utilizing high-performance computing (HPC) clusters. However, the surge in AI compute intensity, which is doubling every six to 10 months, has put significant pressure on data centers to provide the necessary power and cooling solutions.
Giordano Albertazzi, CEO of Vertiv, a global provider of critical digital infrastructure and continuity solutions, emphasizes the need for high-density infrastructure to meet the rising demand. High-density cooling and power have become essential as data- and compute-intensive AI workloads require more resources. To address this challenge, companies like Vertiv are focusing on innovative solutions that prioritize technology and sustainability.
The power demands of data center computer racks are projected to increase dramatically in the near future. Omdia’s 2022 Data Center Thermal Management Market Analysis report predicts that typical data center racks could require 50 kW or more, compared to the current 5 kW to 7 kW. This means there is a pressing need to develop and deploy solutions that can deliver ever-increasing power while also improving efficiency.
To tackle the issue of cooling supercomputers, different approaches are emerging. One example is the use of rear-door heat exchangers combined with air cooling, which can effectively displace heat outside the servers. Another option is immersion cooling, which submerges servers and components in a thermally conductive liquid. Additionally, direct-to-chip liquid cooling involves delivering cooling liquid through cold plates on top of heat sources, drawing away the heat as it circulates.
In the future, data centers will likely adopt a hybrid approach, tightly orchestrating both liquid and air cooling to optimize their operational environments. The evolving demands of AI necessitate constant problem-solving from teams of engineers, as both existing and future data centers seek to retrofit or design facilities that can provide the required power and cooling densities.
As AI continues to reshape various industries, its true potential relies heavily on data centers’ capacity to support its computational intelligence. The industry must continually evolve to offer dynamic and innovative cooling and power solutions that not only address the challenges posed by evolving data centers but also maximize the potential of AI.
In conclusion, the growth and reliance on AI, particularly GenAI, is driving the need for data centers to adapt to increasing power and cooling demands. Companies like Vertiv are at the forefront of developing solutions that balance technology and sustainability. Through innovations such as alternative energy sources and smart data center designs, the industry aims to meet the rising power requirements while minimizing their environmental impact. It is clear that the future of AI is closely intertwined with the evolution of data center infrastructure, making it crucial to address the power and cooling needs to unlock the full potential of AI applications.
