Sustainability is the word on all cloud executives’ lips right now. Yet whilst the consensus on the need for urgent action is clear — especially given the digital sector’s 4% contribution to global warming — the divergence of solutions currently proposed is just as striking.

This cacophony was particularly clear at the 2022 edition of CloudFest, held at EuropaPark March 22-24. Whilst sustainability dominated the conference programme, no one could agree on a single path to take.

For chipmakers like Ampere, more efficient CPUs are naturally the way forward. Ampere’s "cloud native processor" is nearly twice as energy-efficient as AMD Milan, claimed the company’s CPO, Jeff Wittich. This, and having 2.4x more cores per rack on their CPUs than the competition, is the key to reducing cloud computing’s carbon impact, according to Ampere.

For manufacturers such as Lenovo, the best results can be had in the supply chain domain (or "how do we ship more in less space?", as Alexander Jenewein, the firm’s EMEA Hyperscale Sales Leader put it), but also by making PCs out of recycled plastic… and by offering customers a carbon offset service.

As for cloud companies, Ionos’ Head of TechOps Infrastructure Stefan Mink also concentrated on the packaging aspect, not just for shipping servers, but also for making casings from recyclable plastic, and for ensuring chassis are as adaptable as possible, in order to optimize rack space.

Meanwhile, “cloud PaaS” company suggested ‘geo-relocation’ could be a leading solution. Namely moving Ireland-based workloads to France or Sweden, where carbon intensity is 10 times better. Or, similarly, relocating US workloads to France, which could reduce emissions by up to 90%, said the company’s Mathieu Strauch.

Some more concrete options came from the “Cloud Sustainability And Clean Energy Made Simple” panel, which featured Roger Benson, Commercial Senior Director, EMEA at AMD; Mike Israel, IT Asset Lifecycle Consultant at Hewlett Packard Enterprise; and Yann-Guirec Manac’h, Head of Hardware R&D at Scaleway.

For AMD as for Ampere, the answer lies in creating “the best building blocks for servers (CPUs) that are extremely efficient.” This has notably been the case with Scaleway, where collaboration with AMD on a new type of configuration has reaped great rewards, said Manac’h “It’s a perfect example of ideal resource management”, he said. “Going from four socket to single socket server a few years ago, everyone told us we’re mad! But it’s been a great success.” Benson concurred that “our clients really see the advantages of our CPUs in terms of fewer servers and significantly less energy consumption. We’ve been going from strength to strength with Scaleway and other cloud providers (CSPs) since then. If you’re a CSP, you want to be transparent about your sustainability goals, in order to communicate clearly with your clients, and reduce your own carbon footprint. I don’t think we’ve gone nearly far enough in this domain yet.”

The panel above all agreed that equipment reuse was a key way forwards in terms of reducing environmental impact. “The lifespan of a server is 3-4 years. That’s not great”, said Manac’h. “We’re trying to push the life of our equipment up to 10 years. 15-30% of a server’s carbon impact comes from its manufacturing. So you can reduce a lot of carbon impact — and reduce e-waste — by reusing your equipment. The (chip) crisis for us means we have to change our way of thinking. Reusing equipment makes even more sense now than before. Alternatively, it’ll all pile up, and we’ll end up living in serverland!”

Indeed, added Manac’h, Scaleway’s current objective of repurposing at least 13,000 servers — which is currently 26% reached — will ultimately save the equivalent of 7200 tonnes of CO2, according to Dell’s own calculations of the carbon impact of server manufacturing. That’s the equivalent of 5000 Western European households’ annual electricity consumption.

HPE’s Israel also insisted his company “takes care of old equipment. We take it back, and remarket 85-90% of it. The circular economy is key”, he added. “We support this as far as possible. Then when equipment is unusable, we recycle it all. In the end, only 0.5% of it is e-waste.”

Reuse can also happen on a tiered, global scale; a process Benson called ‘cascading’. “We’ve seen some companies cascade their unused machines to their affiliates in other territories that are less advanced. It’s good to see that sort of thing,” he said.

The chipmaker remains, however, resolutely set on reducing its CPUs power consumption, as proves its latest “30 by 25” drive, which aims to multiply efficiency by 30 by 2025, by reducing power consumption per unit by over 97%.

Looking ahead, the panel concluded by trying to predict future trends, which will most likely involve reconciling surging data usage with the need to limit resource consumption. Whilst AMD is looking to constrain its own products’ energy usage, HPE’s Israel predicted that more and more data “would be produced at the edge. Look at autonomous driving, you can’t have a central DC for that. Rather, you need micro-, or decentralized data centers.”

“You want your data to be processed fast, be it on the cloud, on the edge,” countered Manac’h. “You just need it to run, you don’t care where. If every car on the road is autonomous, there’s no way we can crunch all that data with the infrastructure we currently have.”

Could reused equipment be one way to help handle the surging demand for cloud resources whilst at the same time limiting its impact on the planet? Let’s hope so…