BBC TV's Panorama’s programme that aired on the 6th February would lead some to conclude that cloud computing stands in opposition to environmental sustainability. Their focus was consumer applications such as Facebook TikTok or gaming – and the ‘greedy’ data centres that support these apps, which seem so essential to modern life. However, there is a strong pro-sustainability argument for the cloud.
Cloud typically is more sustainable than alternatives
The independent academic literature is sparse but a review of other studies by Masanet et al. in 2013 indicated that a typical carbon saving of moving computing resources from on-premises data centres to the cloud is 87%. More recent industry research from 2020 settled on an 84% reduction. This is because cloud providers allow multiple customers to use the same underlying hardware, ensuring utilisation is optimised. Many on-premises devices do not function in the same way – for example, a colleague of mine recently discovered a router in a client’s data centre that had been plugged in for years, but its connection to the rest of the network had been physically severed. Moreover, cloud providers have optimised the cooling in their facilities so that the power usage efficiency (PUE) is generally greater. PUE definitions vary (event ISO/IEC 30134-2 has limitations) - but it attempts to capture the ratio of data centre power usage vs that which is used to power computing devices.
Panorama did highlight a side effect of the move to the cloud, which is to concentrate power (and sometimes water) demands in a smaller area. In recent years, Dublin’s data centres have come under scrutiny for the load they are putting on the electricity grid – some of that load would have previously been spread out across the UK and Ireland in smaller, less efficient data centres. This has real knock-on impacts – for example, Panorama described housing schemes being shelved because the local grid cannot cope with the extra load. However, these short-term pains must be weighed against the longer-term environmental benefits.
Cloud providers are addressing sustainability rapidly
The major cloud providers all have ambitious commitments on the use of renewable power, and some extend to water used in cooling. We partner with all three, and while the levels of commitment vary slightly, both Microsoft Azure and Google Cloud Platform will use 100% renewable energy 100% of the time by 2030. Although Microsoft and AWS will be using 100% renewables by 2025 (Google achieved this in 2017), these commitments are achieved through Purchase Power Agreements (PPAs), which means that an amount of renewable energy is logically allocated to each firm – not that data centres are running 24/7 on renewables.
The 2030 commitments are more onerous. These kinds of ambitions go beyond most other industries, and IT can innovate rapidly, making the pace of change is incredible. On water, Microsoft have committed to replenishing more water than it consumes by 2030. New solutions are also emerging for on-premises data centres – for example, our partner QIO can reduce the clock speed of servers when not in use – which has a direct benefit on energy saved for the computer itself and indirect gains through reduced cooling. Cooling itself is also becoming more intelligent, thus creating lower impacts.
What isn’t going fast enough is reporting for end customers on the carbon impacts of their public cloud workloads (across scope 1, 2 and 3). While Google offers a breakdown by application (project), Azure does not offer the same detail, and AWS lacks scope 3 emissions entirely. A lack of graduality or coverage undermines the commitments described above because it becomes challenging to trade off public cloud against on-premises alternatives. This is crucial because we need to put carbon data in the hands of those making investment decisions. Behind the bold stats at the start of this blog will be scenarios where an on-premises location is already optimal for renewables directly from the grid (rather than PPAs), already has low energy cooling and is using bare-bones hardware which reduces embedded carbon. In these cases, an on-premises solution could well be the lower emissions option – the trouble is, it’s hard to be definitive about specific scenarios.
Cloud IT as a decarbonisation solution
While there is a great deal of hype around the use of IT solutions to promote sustainability in other industries (every vendor has a story), there are also genuine attempts to shift the dial. Just one example I’ve recently been involved with is in the agriculture sector, which is responsible for 18% of global emissions. Changes are necessary on individual farms such as alternative land use and greater adoption of solar energy, but many farms are small and already operating on low margins.
We are working with our client to gather data about farm sustainability measures which can drive a better price for produce. Lower emissions in the supply chain is important to big supermarkets, meaning there are potential financial benefits. IT platforms bring this data together to drive system-wide changes. For example, digital transformation is already significantly reducing the need for human travel and eliminating paper. However, in that process, some carbon is created both in the development and operation of IT platforms.
Not every use case has this positive trade-off. Panorama ended up highlighting two very different uses for TikTok’s cloud – trivial (think cat videos) and serious (raising awareness about water shortages in the Colorado River, for example). I don’t think that irony was lost on the presenter.
Optimal use of cloud
Although the cloud model is inherently more efficient than alternatives, like any resource, it can be used ineffectively. Organisations might have good system architecture reasons to ‘reserve’ compute capability, but we also see plenty of inefficient cloud configurations. We typically find 30%-40% of cloud capacity is wasted – which is enough for NTT DATA to run a business line around cloud optimisation, funded entirely through sharing the savings with our client.
Part of the reason this problem exists is cultural, so we work with customers to help change behaviours and optimise technology patterns across the supplier ecosystem, so there are lasting financial and carbon savings. However, optimisation starts at the software design stage (or even before that), and the change we are driving with the Green Software Foundation helps ensure that best practice is spread worldwide.
No get-outs for the IT industry
Having said all the above, there is no escaping from the fact that the more compute power we use, the more there is an impact elsewhere. Unless hyperscalers are generating their own electricity, the use of renewables in IT means less for others (although, as some of the biggest purchasers of renewable energy, their demand signals already help drive the market for clean energy). Moreover, there is a carbon footprint involved in creating IT hardware, building data centres and connecting them to communication infrastructure. These are all reasons for the IT industry not to be complacent. NTT is innovating in this space already – for example, we are backing IOWN, a technology which replaces electricity with light for communications, at a fraction of the environmental impact.
Cat videos
Panorama concluded by turning the spotlight back on consumer demand for digital entertainment services as a driver of data centre growth – of course, it was thanks to the BBC’s iPlayer that I was able to watch the programme at all. While the impacts of air travel, petrol cars and food miles are becoming better understood by consumers, this is scratching the surface of how our lifestyles affect the planet. In addressing this issue, we may find a different challenge emerges: the rebound effect. As hyperscalers make progress on emissions reduction and energy usage drops, the financial and carbon costs of the cloud also reduce, paradoxically driving more cloud consumption. Since we’re a long way from truly zero-carbon cloud computing, hard choices lie ahead if the work being done both in data centres and in software design to reduce the environmental impact of the cloud is not to be supplanted by the “cat videos” use case of the future.
This article has been edited since first being published to address helpful feedback from the wider community.