The Race for Sustainability in the HiPEAC Vision 2024  – Communications of the ACM

For some years, the HiPEAC Vision has focused on sustainability as a key concern for the computing systems community. In discussing this topic, this article summarizes the conclusions of papers by Koen De Bosschere, Patrick Blouet, Thomas Ernst, Lars-Åke Ragnarsson, Jean-Pierre Raskin, and Lieven Eeckhout, all of which form part of the rationale for the HIPEAC Vision 2024.

‘The greatest shortcoming of the human race is our inability to understand the exponential function’ – Albert Allen Bartlett

Arguably the greatest challenge of the 21st century, sustainability is a dauntingly large and multidimensional problem. Any sector, including that of information and communication technology (ICT), is only part of a much wider landscape. Understandably, computer engineers and computer scientists may throw up their hands and think that this is not their problem to solve.

Credit: Roger Castro/Monzón

However, there are some compelling reasons why those of us in the sector should at least attempt to address the sustainability question. Take, for example, the fact that the carbon footprint of the ICT industry has been rising by 1.8% per year over the last few decades; currently it is estimated to contribute between 2.1% and 3.9% of global greenhouse gas emissions – which, as Eeckhout notes, is on a par with the aviation industry [1] – and this figure is set to rise. The ICT sector as a whole, meanwhile, is estimated to have represented 9.3% of global energy use in 2021 and to reach 11.5% of global energy use by 2030 [2].

Or consider the fact that virtually every element of Mendeleev’s periodic table – bar radioactive materials – goes into the circuits used in smartphones. Yet leading electronics recycling companies have only managed to separate 17 of these 60 different materials for recycling [3]– and that’s before you take into account that only 17.4% of electronics waste is currently recycled at all [2]. The colossal water use of the ICT industry is another major concern: as an example, a semiconductor fab today is estimated to use around eight liters of ultrapure water per square centimeter of wafer. In other words, a fab producing 20,000 wafers of 200 mm per month would need more than one Olympic-sized swimming pool of water per day [2].

On a shared planet of finite resources, statistics like these should be enough to give pause to anyone working on ICT products and services. Certainly, we felt that sustainability should be given a major focus in the HiPEAC Vision 2024, with a dedicated section featuring contributions from experts in the field. The starting point, the working definition of ‘sustainability’, as cited by Eeckhout in [1], is that of the 1987 Brundtland Report of the World Council on Economic Development: “Sustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs.” [4]

Credit: Arnout Fierens

The life and death of electronic devices

The conversation about ‘sustainability’ in computing often starts and ends with energy efficiency in the operation of devices and infrastructure. While clearly part of the puzzle, this does not take into account other aspects of sustainability, such as material use and water consumption. As noted above, the majority of non-radioactive elements on the periodic table are used in integrated-circuit manufacture, and these are obtained via extractive, linear mining processes rather than being derived from recycling. What is more, due to process steps involving sacrificial layers, a large amount of material is wasted in comparison to the silicon that ends up in the final chip. Manufacturing processes also use gases, chemicals, precursors, metals, and ultrapure water [3]. Water use in the ICT sector is also becoming increasingly contentious in a world where many areas experience water scarcity: TSMC had to abandon plans for a 2 nm factory due to protests in drought-ridden Taiwan, while Microsoft had to change its datacenter construction plans to comply with environmental restrictions in Arizona [2].

Even if we limit the focus to energy, it’s important to understand that the energy used in the production of devices—that is, the ‘embodied’ energy of a device—is relatively greater than the energy used while the device is in operation, i.e., the ‘operational’energy. Embodied emissions relate to, in Eeckhout’s words, ‘raw-material extraction, manufacturing, assembly, transportation, repair, maintenance and end-of-life processing’ while operational emissions relate to ‘product use during a device’s lifetime’. As Eeckhout notes, embodied emissions already are, or soon will be, the biggest contributor to greenhouse gas emissions in the ICT sector, which he attributes to the increasing demand for chips and the growing energy intensity of semiconductor manufacturing [1].

While energy-efficient devices may mean that the operational energy is reduced, the embodied energy required to produce a device may mean that it switching to it does not lead to a sustainable outcome overall. One example cited by De Bosschere and Blouet is solid-state disks (SSDs), which consume less power than hard disks, but which have an embodied carbon footprint that is eight times greater than that of hard disks [2]. Contrary to popular belief, the route to using less energy overall may well be to prolong the lifetime of older, less energy-efficient versions of a particular technology.

Credit: Arnout Fierens

Hence the drive for more energy-efficient devices, while very welcome, is not sufficient to achieve sustainability in the ICT sector. As De Bosschere and Blouet put it, “efficiency is not a proxy for sustainability,” partly due to the rebound effect known as Jevons’ Paradox [2]. Named after the nineteenth-century economist W.S. Jevons, this is the observation that technological improvements leading to greater efficiency result in greater overall consumption of a particular resource (coal, in Jevons’ case). Efficiency in the ICT sector has led, for example, to the addition of more features and the production of more sophisticated chips which, in turn, leads to higher energy use in the sector as a whole. As De Bosschere and Blouet note, while the ICT industry should reduce its greenhouse gas emissions by 50% between now and 2030 in order to contribute its share to climate efforts, in fact total energy consumption of the digital economy is projected to grow by 50% by 2030 [2]. Given that adding a datacenter is the equivalent of adding an extra town to the grid [2], countries wishing to meet their obligations under the Paris Accords of 2015 will have to consider their priorities very carefully.

Finally, in addition to the energy, raw material, and water use during the production and operation of a device, there is the question of what happens to it at the end of its lifetime. In 2019, the world produced over 53 million tons of electronic waste [5]; De Bosschere and Blouet note this is projected to increase to 75 million tons by 2030 [2]. Recycling electronics is very difficult, due to the combination of elements making up electronic circuits, and itself currently requires aggressive acids and solvents [6]. As Ernst et al. note, this “leads to a double penalty for the global south countries,” which have to deal with the environmental pollution resulting from raw-material extraction, as well as receiving 75% of e-waste from higher-income countries [3], which ends up leaching toxic chemicals into the environment.

Paths to sustainability

Clearly, then, a genuine (as opposed to greenwashing) attempt to grapple with the sustainability issue in ICT will involve revisiting the way devices are currently produced and used. The starting point should be reliable data, without which it is impossible to make informed decisions. Standardized benchmarks – such as those developed by imec’s Sustainable Semiconductor Technology and Systems (SSTS) program [1]– would also help to assess progress by different players in the industry. Once data is available, this should be used to create lifecycle assessments and models, without which, as De Bosschere and Blouet point out, “it is impossible to explore the design space for the most sustainable solution.”[2].”

Sustainability-focused design methodologies will also need to be developed: do larger chips have a larger embodied footprint? (Spoiler: yes [1]). Can we create accelerators in a more sustainable design paradigm? Which microarchitectures deliver higher performance with a lower environmental footprint? Do low-level, natively compiled languages contribute to a lower operational footprint than high-abstraction, managed programming languages? (All these issues are discussed by Eeckhout in [1]).

Production techniques also need to be adapted for a sustainable outcome. Bio-waste-based chemicals and materials for the lithography process can reduce the use of solvents and chemicals, while bio-based materials could also be used for packaging; research into these areas is currently ongoing [3].

To make all of this happen, business models need to be reworked to incorporate the full lifecycle and to account for other values than just profit. Legislation will likely be required to accompany and shape this shift.

Finally, it is perhaps time that we started prioritizing the use of ICT, itself a precious resource. The most beneficial use cases, including ones that power the shift to a sustainable economy respecting planetary limits, should be prioritized. After all, sustainability is non-negotiable: sooner or later, a physical limit will be imposed, whether it is minerals becoming unavailable, an inability to produce sufficient energy, water supplies running out, or the catastrophic effects of climate breakdown. It behooves the computing community to act before that point is reached.

Credit: Roger Castro/Monzón

The HiPEAC project has received funding from the European Union’s Horizon Europe research and innovation funding program under grant agreement number 101069836. Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them.

Koen De Bosschere is a professor of computing systems, leader of the computer systems lab and head of the electronics and information systems department at Ghent University, Belgium. He is the coordinator of HiPEAC, the premier network of computing systems researchers in Europe, co-editor of the bi-annual HiPEAC Vision, and organizer of the yearly ACACES summer school. Madeleine Gray, HiPEAC communications officer, has specialized in science communication since 2014. She has worked on a number of strategic computing European projects, including EUDAT and the European Processor Initiative, as well as HiPEAC.