Unsustainable Matrix — Digital Infrastructure
Digital Power Consumption Vs. Global Power Supply
With the help of a few colleagues, we pulled together the most credible data sources we could find and created a simple graph.
The graph is based on the following data;
- Total Global Electricity/Power Supply — w/ growth rates to 2030
- Total Global ICT (mostly data centers) Electricity/Power Consumption — w/ growth rates to 2030
This is what it looks like:
And here is the same graph, with each power consumption scenario shown as a percentage of available power:
To unpack it a bit further:
- We used/assumed the medium growth rate scenario for power supply growth.
- The ICT power consumption scenarios vary primarily based on an assumed rate of energy efficiency growth over the time period.
- The low scenario assumes a high rate of efficiency growth achieved. The “expected” case as per Anders Andrae is the medium ICT scenario.
- The magenta colored line is the “medium” case for electricity supply growth, per the graph below;
As a disclaimer; sources for this data are limited, but are as credible as we could find. There are plenty of gaps in the data and I expect these will fill-in over time.
The purpose is simply to illustrate the likely power consumption growth curve of digital infrastructure, relative to the growth curve of the power supply globally.
The growth of IP traffic/cloud storage is driving the growth of data centers (servers), and these data centers are huge consumers of power/electricity.
And for an added variable, likely the driver for the projected growth, we also graphed the global IT traffic/storage growth:
Further questions stem from this graph, as it is not clear whether the full extent of digital/data demand is captured here.
The convergence of cloud, 5G, crypto-mining, blockchain, IoT, AI/ML, autonomous driving, and related digital demand is difficult to fully unpack.
That being said, I think it is worth simplifying and publishing as the implications of even the “high-efficiency, low scenario” are considerable and only 10 years away.
More to come on this, along with further questions, for example;
- What does this look like in 2040? 2050?
- How are we going to power this infrastructure sustainably?
- Can utilities/grids support this? Is it too risky to rely on the utility?
- Do private power plants (micro-grids, fuel cells, storage, solar) make more sense from a risk perspective? Are they more financeable with such credit worthy off-takers?
- What are the long term risks to relying on utility power for such power-dense infrastructure?
Comments/Questions appreciated
Note: digital infrastructure refers to the physical hardware/systems that enable digital communications, cell towers, the internet, cloud, IoT, blockchain, and everything related to IT.
Sources:
- Cisco (2018b) Cisco Visual Networking Index: Forecast and Trends, 2017 2022 White Paper
- ITU (2015) IMT Traffic Estimates 2020–2030
- Andrae (2017) Total Consumer Power Consumption Forecast
- IEA (2019) Data centres and data transmission networks
- McKinsey (2019) Global Energy Perspective 2019: Reference Case
- Shi, Pallis, Xu (2019) Edge Computing [Scanning the Issue]
- BP (2019) BP Statistical Review of World Energy. June 2019. 68th Edition
- EIA (2019) US Energy Information Administration — EIA projects 28% increase in world energy use by 2040
- Andrae (2015) On Global Electricity Usage of Communication Technology: Trends to 2030
- IDC (2018) The Digitization of the World From Edge to Core