E D I T O R ' S Q U E S T I O N

For this month ’ s Editor ’ s Question , we receive insight from data centre experts sharing how operators can manage the requirements for uptime while reaching targets for a greener , more sustainable future .

“ Global demand for data centres and computing power has exploded since the expansion of GenAI and LLMs , pushing companies ’ requirements to new heights as management teams accelerate their implementation of AI capabilities .

“ At the same time , a swathe of new rules such as the EU ’ s Corporate Sustainability Reporting Directive and the SEC ’ s rules on climate-related disclosures will come into effect in the coming years , seeking to quantify and arrest carbon emissions , while many data centre operators also have their own net zero goals to meet .

“ On the face of it , meeting expanded demand while managing and mitigating the impacts of a business on the environment may seem to be competing demands .

“ Resilience requires having excess capacity to meet increased demand ( on underlying services and utilities ) while complying with sustainability regulation involves the mitigation of excess emissions over time . Data centre operators are therefore being pulled in two directions as they seek to enhance their capacity and resilience without compromising the push to meet energy and carbon reduction objectives over time .

“ Historically , the drive for commercial resilience trumped sustainability considerations , with minimal automation leading to an over-reliance on operationally intensive methods , such as manual monitoring and service management activity . This resulted in high infrastructure redundancy levels and replication of data and backups . But now there is an increasing shift in priority from the end-users of data centre services , and resilience and sustainability no longer need to be mutually exclusive .

“ Cloud-native infrastructure and application architecture patterns prioritise automation , scalability and recovery mechanisms , so applications can be lightweight and scale up and down proportional to demand . In addition , there is an added commercial incentive to adopting Financial Operations ( FinOps ) principles , which usually have a corresponding energy consumption reduction benefit .

“ AI has been made possible through LLMs requiring access to more and more compute to train models , whereas , for a large subset of AI use cases , endusers may find Small Language Models ( SLMs ) a more effective , efficient , secure and eco-friendly approach to solving problems .

“ Data centres are already estimated to contribute 2 – 3 % of global greenhouse gas emissions . This will only grow with the rise of increasingly large AI models , and until new methods of computation can drive orders of magnitude efficiencies , for example through Quantum Computing . Until then , a more sustainable approach to operating data centres will be an environmental and economic priority .”

www . intelligentdatacentres . com 31