Lost in Migration? Attributing carbon when outsourcing to data centres and cloud
This discussion paper is the outcome of a long running dialogue between techUK and members of HMG’s Sustainable Advice and Reporting Team, (STAR). We were joined by several large media organisations and two carbon footprinting service providers. Government and media organisations are large customers of cloud service providers and both reported that they were struggling to understand the energy and carbon implications of these outsourced services. Our discussions therefore focused on how feasible it is to attribute carbon to cloud services and what we can do collectively to improve energy and carbon transparency in these processes.
When organisations outsource digital activity to third party data centres or cloud service providers, there is a general assumption that efficiency improves, and that energy and carbon benefits are realised. But how do we know that this is true? When companies manage their own services, whether analogue or digital, they have access to relevant energy use data- at least in theory – firstly because they pay the bills, and secondly because they control the processes so they can attribute consumption appropriately. But when a service is outsourced to a third party the energy impact of that activity sometimes becomes much less transparent. So how can organisations demonstrate, with a reasonable degree of confidence, that their outsourcing decision is indeed delivering environmental benefits – or even more importantly, identify occasions when it is not?
This document explores how we can go about attributing carbon to the activity we outsource to third party data centres and providers. The objective is not to advocate cloud adoption or provide a calculation methodology, but to identify ways in which customers can understand these impacts, or at least estimate them robustly enough to inform their decision making. It represents the outcome of our discussions but in reality it is the start of a conversation, not a finished product.