Paper notes – Moore’s Law and ICT Innovation in the Anthropocene

I get Google Scholar alerts for new publications searching against a few different keywords related to sustainable computing. This usually emails me 2-5 new papers each week, of which maybe 1-2 are relevant / sound interesting. I aim to read one per week, take notes, make highlights, and try to stay up to date with the latest research.

I thought it may be useful to publish my notes as I go, so this will be an ongoing series.

Paper #

Bol, D., Pirson, T. & Dekimpe, R. (2021) Moore’s Law and ICT Innovation in the Anthropocene. Available from: https://dial.uclouvain.be/pr/boreal/object/boreal%3A243578/datastream/PDF_01/view

Abstract #

In information and communication technologies (ICTs), innovation is intrinsically linked to empirical laws of exponential efficiency improvement such as Moore’s law. By following these laws, the industry achieved an amazing relative decoupling between the improvement of key performance indicators (KPIs), such as the number of transistors, from physical resource usage such as silicon wafers. Concurrently, digital ICTs came from almost zero greenhouse gas emission (GHG) in the middle of the twentieth century to direct annual carbon footprint of approximately 1400 MT CO2e today. Given the fact that we have to strongly reduce global GHG emissions to limit global warming below 2°C, it is not clear if the simple follow-up of these trends can decrease the direct GHG emissions of the ICT sector on a trajectory compatible with Paris agreement.

In this paper, we analyze the recent evolution of energy and carbon footprints from three ICT activity sub-sectors: semiconductor manufacturing, wireless Internet access and datacenter usage. By adopting a Kaya-like decomposition in technology affluence and efficiency factors, we find out that the KPI increase failed to reach an absolute decoupling with respect to total energy consumption because the technology affluence increases more than the efficiency. The same conclusion holds for GHG emissions except for datacenters, where recent investment in renewable energy sources lead to an absolute GHG reduction over the last years, despite a moderate energy increase.

We formulate hypotheses for this absence of absolute decoupling from three scientific fields: ecological economics, economics of technology and sociology of technology. We argue that aligning direct GHG emissions of the ICT sector on a trajectory compatible with Paris agreement requires an ecological transition in innovation by adopting sobriety in addition to efficiency.

Notes #

• This paper draws on 6 studies which examine global ICT footprint. They summarise the findings as: annual electricity consumption and GHG emissions of approximately 1400TWh and 1400MTCO2e for the ICT sector today. However, only 2 of those studies have received peer review (which the authors note in the table) and there is no discussion of the definition of “ICT”, which may vary across the studies.
• Energy consumption is easier to estimate because it can be measured. Annual GHG emissions from ICT tend to be linked to energy consumption but they are only a snapshot because the emissions factors related to energy consumption change over time. The authors note this later in the paper in relation to data centers and the investments in renewable energy that powers them.
• The authors note that for GHG emissions to fall, we need to consider the full lifecycle. This varies depending on the item you are measuring, such as servers, which have large use-stage emissions vs user devices which tend to have large manufacturing emissions.
• The key question of the paper is whether efficiency improvements will be helpful to reduce emissions, and in particular whether the expected ICT usage increases are decoupled from the carbon emissions associated with that usage. Can we reduce carbon emissions as we use more and more ICT services?
• Semiconductors: Transistor density is increasing as dimensions reduce, from 90nm down to the current cutting edge 5nm sizes. These smaller transistors are much more complex to manufacture and the energy required in the processes has increased by 12-20% annually. Semiconductor manufacturing is probably the most complex industrial ever created, so it is not surprising that chip production makes up a large share of the energy and carbon footprint of electronics manufacturing.
• The paper also considers the physical limits the ICT industry is rapidly approaching, for example semiconductor transistor sizes cannot go below the size of a single atom. What happens when we reach that limit and cannot make chips more efficient?
• Mobile internet access: Is growing rapidly, particularly because developing countries are skipping many of the stages of telecoms deployment that more developed countries have gone through. More people are coming online and they are using 4G mobile technology to do so. 5G will grow rapidly over the coming years, but despite the potential for energy efficiency improvements, the ability to transmit more data, faster i.e. more efficiently, that doesn’t mean total energy consumption will decrease.
• Data centers: Efficiency improvements over the last 10 years have been enough to offset increased energy consumption so that the total growth estimates range between 0.7% and 6.5% per year. However, this appears to be slowing.

Conclusions #

The paper concludes that there is no evidence to support a decoupling of carbon emissions from economic growth on a global scale, but highlights data centers as an interesting example because that sector has shown it is possible (depending on which estimate you look at). This improvement is mainly driven by the hyperscale providers who are purchasing large amounts of renewable energy, but the paper notes that this may not scale to other industries:

..if we consider a non ideal market which is likely to be the reality, the low-carbon energy purchased by datacenter operators may no longer be available for other sectors. In this case, the decoupling remains local and limited GHG reduction is achieved at the global scale because the energy footprint is not decreasing.

Further, as we make more efficient equipment, there is a tendency to throw away old items to get the new shiny thing. This may make sense if the majority of the emissions come from the use-stage, but that is not the case for consumer devices where this forced obsolescence tends to manifest.

The paper makes an important conclusion that focus on ever more efficient technologies is not going to help us reduce our carbon footprint. This is the classic example of the Jevons Paradox.

Finally, the paper discusses the concept of “speculative faith in technology”:

We thus define KPI-driven innovation as the blind habit for ICT engineers and companies to pursue the increase of any KPI such as number of CPU cores, number of artificial neurons, number of pixels in a CMOS imager, number of imagers in a smartphone, with the unconscious faith that it will lead to future valorization. Secondly, disruption of the present by technological innovation is a key speculative practice to invent a future that is supposed to ”change everything” and absorb the debts of the present.

Whilst it is true that this kind of innovation for the sake of innovation buzzwords is not ideal, this argument is not developed further. It sounds like a criticism of the cliche Silicon Valley startup parodied in the mainstream press rather than a specific criticism backed by examples. There will always be people chasing the latest shiny things – it’s impossibe to know in advance what will and won’t work. Their argument that this is a negative shows a misunderstanding of the startup process, and how technological innovation works. I’m not sure why this was included in an otherwise interesting paper.