Hypotheses for Primary Energy Use, Electricity Use and CΟ2 Emissions of Global Computing and Its Shares of the Total Between 2020 and 2030

Abstract

There is no doubt that the economic and computing activity related to the digital sector will ramp up faster in the present decade than in the last. Moreover, computing infrastructure is one of three major drivers of new electricity use alongsidefuture and current hydrogen production and battery electric vehicles charging.
Here is proposed a trajectory in this decade for CO2 emissions associated with this digitalization and its share of electricity and energy generation as a whole. The roadmap for major sources of primary energy and electricity and associated CO2 emissions areprojected and connected to the probable power use of the digital industry. The
truncation error for manufacturing related CO2 emissions may be 0.8 Gt or more indicating a larger share of manufacturing and absolute digital CO2 emissions.While remaining at a moderate share of global CO2 emissions (4-5%), the resulting digital CO2 emissions will likely rise from 2020 to 2030. The opposite may only happen if the electricity used to run especially data centers and production plants is produced locally (next to the data centers and plants) from renewable sources and data intensity metrics grow slower than expected.

Full text

Hypotheses for primary energy use, electricity use and CO2 emissions of
global computing and its shares of the total between 2020 and 2030
ANDERS S.G. ANDRAE
Huawei Technologies Sweden AB
Skalholtsgatan 9, 16494 Kista
SWEDEN
anders.andrae@huawei.com
Abstract: - There is no doubt that the economic and computing activity related to the digital sector will ramp up
faster in the present decade than in the last. Moreover, computing infrastructure is one of three major drivers of
new electricity use alongsidefuture and current hydrogen production and battery electric vehicles charging.
Here is proposed a trajectory in this decade for CO2 emissions associated with this digitalization and its share of
electricity and energy generation as a whole. The roadmap for major sources of primary energy and electricity
and associated CO2 emissions areprojected and connected to the probable power use of the digital industry. The
truncation error for manufacturing related CO2 emissions may be 0.8 Gt or more indicating a larger share of
manufacturing and absolute digital CO2 emissions.While remaining at a moderate share of global CO2
emissions (4-5%), the resulting digital CO2 emissions will likely rise from 2020 to 2030. The opposite may
only happen if the electricity used to run especially data centers and production plants is produced locally (next
to the data centers and plants) from renewable sources and data intensity metrics grow slower than expected.
Key-Words: -carbon dioxide, data centers,electricity, devices, hydrogen,manufacturing, networks, primary
energy, production, renewables, steel.
1 Introduction
There are reasons to believe that global primary
energy consumption (GPEC) will increase in this
decade. For instance, U.S. Energy Information
Administration (USEIA) estimates that GPEC will
increase by 28% between 2015 and 2040 and that
Coal, Oil, Gas will make up more than 75% of
GPEC in 2040 [1]. USEIA also predicted that global
energy-related CO2 emissions will grow 0.6% per
year from 2018 to 2050 [2]. In summary, the global
demand of Coal, Oil, Gas is expected togrow
towards 2050[3]. Here it is hypothesized that
between 2020 and 2030 the global CO2 emissions
related to energy conversion increase >10% from
≈36 Gigatonnes (Gt) to ≈40 Gt, i.e. 1% per year. 1%
is consistent with the growth rate for global energy
related CO2 emissions reported in [4] from 2007 to
2017.
The motivation for repeating the global energy
situation is to put the entire digital sector
(hypothetically having similar scale of power use
asall kinds of computing/processing) into
perspective being ≈3% of GPEC, ≈7% of global
electricity use (GEU), and ≈5% of global CO2
emissions.
Table 1 adapted from [5] shows potential trends of
computing power and intensity.Computing intensity
is here expressed in Zettainstructions per second
which is a common metric for computing capacity
[6]. Computing energy efficiency is commonly
expressed in instructions per kWh[7]. Computing is
assumingly a larger entity than the digital sector but
should have similar magnitude as far as power use.
Table 1. Potential historical and future trends for
computing and global power.
Year
Zetta
Instruc
tions
per
second,
ZIPS
[5]
Gigainstr
uctions
per J,
GIPJ [5]
GW
average
power for
computing
[IPS/IPJ]
GW
average
power for
World
(from
Table 4)
Shar
e
com
putin
g
2030 2606 5121 560 4103 14%
2029 1622 3297 541 4023 13%
2028 941 2123 487 3944 12%
2027 557 1367 448 3866 12%
2026 343 880 429 3791 11%
2025 215 567 417 3716 11%
2024 104 365 313 3643 9%
2023 61 235 285 3572 8%
2022 35 151 254 3502 7%
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
50
Volume 15, 2020

2021 22 97 248 3433 7%
2020 13 63 233 3366 7%
2019 8.4 40 230 3300 7%
2018 5.4 26 227 3235 7%
2017 3.4 17 224 3172 7%
2016 2.2 11 222 3110 7%
2015 1.4 6.9 217 3049 7%
2014 0.8 4.5 196 2989 7%
2013 0.5 2.9 197 2930 7%
2012 0.3 1.9 197 2873 7%
2011 0.2 1.2 179 2817 6%
2010 0.1 0.8 146 2761 5%
7% is not far from “conventional
wisdom”concerning the current share of information
processing and computing (excluding wireless and
displays) [8]of GEU.
Fig. 1 shows the graphical display of Table 1.
Fig. 1. Computing power trends in relation to global
power consumption 2010 to 2030.
The globalconsumption by fuel shows that Coal, Oil
and Gas totally dominate occupying>80% of the
GPEC and >60% of the GEU[4]. Such non-
intermittent and easily accessible fuels are currently
the basis for societies moving away from poverty as
they provide power independently of the weather.
Anyway, hydro and nuclear have a big importance
in some nations. It seems very difficult to quickly –
in a few decades - stop the use of Coal, Oil and Gas
considering current ways of food production,
heating and air conditioning. Nuclear seems to be an
effective and efficient way of reducing CO2
emissions, especially if the spent fuel can be reused
as in Gen IV nuclear power systems[9][10]. Solar
power will be important when the large-scale
charging capacity is solved.Sun’s energy supply to
Earth is ≈23000 TW[11], andcurrent GPEC,
≈167000 TWh, is thereforetheoretically covered by
the Sun in just 8 hours.
The digital sector and its infrastructure - belonging
to every activity of daily life - cannot easily be
compared to other sectors. Moreover, different
nations and different companies have totally
different starting points regarding electric power
infrastructure, growth of gross national product and
nature of business growth. The current share – and
future evolution - of direct renewable electricity
supply – beyond the local mix - to all global data
centers and networks is not clear. In 2020 the share
is assumed to be close to zero. When
considering CO2 emissions created by computing,
the locally used grid mix is a significant
factor[12][13]. Global e-Sustainability Initiative
(GeSI) estimated that energy use related CO2
emissions by the Information and Communication
Technology (ICT) sector was 0.8 Gt CO2 in 2019,
rising to over 0.9Gt in 2030 driven by the growth of
the sector, especially by electricity use of the
transmission networks[14]. The scope for ICT used
by GeSI is not evident and likely the 2030 CO2
emissions will be higher than 0.9Gt as shown in the
present prediction. Furthermore, it is not clear how
much the digital sector can influence the global
emissions, anthropogenic and others. Similarly the
Consumer Technology Association reported that 2%
of their member companies released 0.792 Gt in
2017 increasing 3% from 2016[15]. Claims are
often made that the solution is that companies shall
switch to renewables with no critical discussion on
the relation to the development of the power
industry. Still, some ICT companies promise that all
of their data centers will use 100% renewable
electricity before 2030. Such claims need further
scrutiny. In this article some new points will
beraised: the underestimation of the production of
ICT Equipment, hypotheses regarding rising data
center power and end-user devices use. Contrary to
previous studies [16],the current hypothesis is that
data centers use stage and upstream manufacturing
will stand out as the drivers for power use and
related CO2 emissions.
2 Problem Formulation
In the present research the hypothesis is that the
probability is close to 100% that the CO2emissions
from computing and the digital sector will rise
between 2020 and 2030.
3 Problem Solution
The proposed solution to test the hypothesis is to
look forforecasts of GPEC and GEU and sources for
each. Then derive the CO2 intensity for primary
energy and electricity sources. Then summarize
2588ZIPS
5.5TIPS/W
556GW
4190GW
0.00E+00
1.00E+12
2.00E+12
3.00E+12
4.00E+12
5.00E+12
6.00E+12
2010 2015 2020 2025 2030
Computingpower,globalpower,instructionsper
secondandjoule,2010to2030
TIPS IPJ Wforcomputing Wglobalpower
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
51
Volume 15, 2020

forecasts for power use of the major entities of the
computing sector. Then add all data into the life
cycle assessment software tool SimaPro version
9.0.0.31 which facilitates logic connections between
all parameters as well as uncertainty analysis with
Monte Carlo simulation. At last test the probability
of CO2 reduction or increase by combining the CO2
intensities with the energy and electricity uses of
computing.
4 Global energy and electricity use
from 2015 to 2030
The global energy forecasts of GPEC and
GEUare done by several bodies with high validity,
e.g. BP[4] and USEIA[2]. Here is summarized and
analyzed numbers from BP. Energy, electricity and
CO2 emission forecasts and baselines will never be
exact, but trend analyses could be more or less
realistic. BPs statistics for energy, electricity and
CO2 seem to be one of the most realistic.
4.1 Global primary energy consumption
Table 2 shows sources for GPEC for 2015
[17]extrapolated to 2020 and 2030 [4].The
hypothesis used to design Table 2 is that the shares
(%) for all sources - except “Other renewables” - are
expected to grow until 2030 as between 2015 and
2018 and the total GPEC to grow 1.9% per year,
consistent with [3]. Then the remaining share is
allocated to “Other renewables”. “Other
renewables” includes wind, solar, geothermal,
biomass, tidal etc.
Table 2. Sources for global primary energy
consumption (ExaJoule), 2015 according to BP
Statistical Review of World Energy and
assumptions for 2020 and 2030
2015 2020 2030
Coal 158 158 156
Oil 182 205 260
Gas 132 143 169
Nuclear 24 26 31
Hydro 37 42 53
Other renewables 15 29 58
TOTAL GPEC (EJ) 548 602 727
TOTAL CO2 (Gt) 33.5 35.5 40.3
Global GPEC CO2
intensity (Gt
CO2/EJ=kg/MJ)
0.061
0.059 0.055
Table 2 suggests that Coal, Oil and Gas will stay
at >80% of GPEC in this decade.
Table 3 shows typical CO2 intensities assumed to
be consistent with the global CO2 emissions (see
Table 2) and usages of Coal, Oil and Gas.
Table 3. CO2 intensity for sources of primary energy
(Gigatonnes/ExaJoule) 2020
and
2030,
median Min Max
Coal 0.095 0.094 0.096
Oil 0.064 0.06 0.07
Gas 0.053 0.05 0.056
4.2 Global electricity – sources and trends
The hypothesis used to design Table 4 is that the
share (%) for all sources - except “Other
renewables”- are expected to grow until 2030 as
between 2017 and 2018 and the total GEUto grow
2.5% per year. Then the remaining share is allocated
to “Other renewables”. “Other renewables” includes
wind, solar, geothermal, biomass, tidal etc. Toward
2050 solar power will likely increase rapidly and
may provide 27% of GEU [3].
Table 4. Sources for global electricity use in 2017
and 2018 andassumptions for 2020 and 2030
2017[4] 2018[4] 2020 2030
Coal 9806 10101 10450 13098
Oil 870 803 772 671
Gas 5953 6183 6432 8290
Nuclear 2639 2701 2783 3417
Hydro 4065 4193 4342 5466
Other
renewables 2343 2634 2872 4454
25677 26615 27651 35395
Table 5 shows the CO2 intensities for different
sources which are assumed to be consistent with the
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
52
Volume 15, 2020

general understanding of current total global
anthropogenic CO2 emissions from GEU sources.
Table 5. CO2 intensity for sources of electric power
(million tonnes CO2/TWh) 2020 and 2030
2020 and 2030 ,
median Min Max
Coal 0.98 0.67 1.19
Oil 0.89 0.82 0.97
Gas 0.65 0.56 0.79
Nuclear 0.008 0.006 0.01
Hydro 0.006 0.003 0.009
Other
renewables 0.010 0.009
Average global
CO2 intensity
2020 0.543±0.084
Average global
CO2 intensity
2030 0.534±0.082
Tables 2 to 5 show that Coal electricity will rise
from 28% of total CO2 emissions in 2020 to 32% in
2030.
5 Digital sector latest prediction
Given the hypotheses for global energy and
electricity forecasts for this decade mentioned in
section 4, how likely is it that the digital industry
will be able to reduce “its own” CO2 emissions?
5.1 Data centers use
Data centers – now using around 33 GW - are of
large interest as they might use much more power
(≈89 GW) in this decade than in the last. Likely,
both bottom-up and top-down approaches are
required to understand the trends of data center
electricity use in this decade. There are some
circumstantial evidence that we are heading for
much higher global Wattage from computing in the
next decade. A recent study[16]–stated161 TWh in
China alone from data centers in 2018 -suggests that
the global data center electricity use is running
along the expected case in[18]. As a result, globally,
data centers could possibly have used 400 TWh in
2018. Moreover, the current data center networks
using individual fibers are costly, bulky, hard to
manage, and not scalable[19]. Also, a new
hypothesis is that servers currently deliver only
around 30% of their nominal electrical efficiency
improvements as reduced electricity costs and
carbon emissions [20]. Moreover, between 2010 and
2020 Germany’s data center electricity use grew 3%
per year from 10.5 to 14.3 TWh, and 16.4 TWh is
predicted for 2025[21]. Scaling Germany’s data
center intensityper capita to theglobal population
estimate in 2030[22]leads to >2000 TWh for global
data centers in 2030. Furthermore, if in 2030 there
will be around 800-1000 massive hyperscale data
centers (each with around 500 MW capacity)
handling most of the Global data center IP traffic
and they in 2030 run on average 300 MW (not
unrealistic that a 500 MW center will use 60% of its
full capacity in 2030) these data centers will use
>2000 TWh. This suggests a certain massive growth
for global data center electricity use as suggested
by[5][16]. Still, the current hypothesisfor 2020 -
based on updates of [16] -is that data centers will
use around 294±5 TWh emitting 0.16±0.3 Gt CO2.
In 2030 this is assumed to rise to 783±190 TWh
emitting 0.42±0.12 Gt CO2.The effect of 50% of the
data centersusing local renewable power in 2030 -
and 50% global average macro grid power –is
presented in in Section 6.
5.2 Mobile networks use
The mobile sector is generally waiting for
5G.Mobile networks power use may rise from 2020
(≈11GW) to 2030 (≈36GW) butstill not at an
alarming rateas the share of mobile networks of the
whole digital sector will likely still be manageable
by 2030. The currenthypothesis for 2020,based on
updates of [16], is that mobile networkswill use
around 98±2 TWh emitting 0.054±0.08 Gt CO2.
In 2030 this is assumed to rise to 316±130 TWh
emitting 0.14±0.06 Gt CO2. The effect of 50% of
the mobile networks using local renewable power in
2030 - and 50% global average macro grid power–is
presented in Section 6.
5.3 Optical networks use
The forecasts for this decade for fixed networks
were heavily overestimated in [16]. The reason is
that equipment swapping was not considered
appropriately. Optical networks currently use
≈17GW, but that may rise to ≈32 GW in 2030.
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
53
Volume 15, 2020

The reason is that industry is indicating some
concerns for energy efficiency[23]. For Wavelength
Division Multiplexing (WDM) systems, the bitrate
increase so far has been somewhat faster than the
energy-efficiency increase. Therefore, foreseeable
WDM generations tend to consume increasing
power over time despite in the best case using less
than 0.2 nanojoule/bit [23].
The current hypothesis for 2020 based on updates of
[16] is that optical networks will use around 150±20
TWh emitting 0.083±0.02 Gt CO2.
In 2030 this is assumed to rise to 284±140 TWh
emitting 0.15±0.06 Gt CO2.
The effect of 50% of the optical networks using
local renewable power in 2030 - and 50% global
average macro grid power –is presented in Section
6.
5.4Devices use
Devices is a very diverse group of gadgets
consisting of phones, portable computers,
peripherals, smart home devices, Wi-Fi
modems/gateways, and IoT devices.They may use
≈95 GW in 2020 and might decline to ≈89 GW in
line with the most popular hypotheses saying that
the group uses less and less power. The reason for
the popular hypothesis is that a shift topower
efficient tablets and smartphones –ins favour of
laptops and desktops – reduces the overall power
use more than the increased amounts of power
efficient devices shipped and installed.
Here the Wi-Fi modems are moved from the
Fixed Wired networks entity in [16] and added to
this group. Fig. 2 and 3 show the reparation in 2020
and 2030, respectively.
Fig. 2. Shares for devices power use in 2020
Fig. 3. Shares for devices power use in 2030
The current hypothesis for 2020 based on updates of
[16] is that deviceswill use around 830±200 TWh
emitting 0.45±0.12 Gt CO2. In 2030 this is
optimisticallyassumed to decrease to 760±320 TWh
emitting 0.4±0.15 Gt CO2.
5.5Manufacturing processes
Manufacturing of digital equipment will use around
34 GW in 2020 optimistically assumed to slowto
≈27 GW in 2030.
For 2015 it was recently estimated[24] that 1
GtCO2 are emitted upstream for the production of
digital devices including radio,
television, communication equipment and apparatus
as well as computers and office machinery. Thereby
it includes not only the end-user digital devices, but
also digital devices which end up in other user
products like cars, buildings etc). 1 GtCO2 per year
is several times higher than other estimations of ICT
hardware production in 2015 [16].
However,[25]also identified manufacturing as being
underestimated, but not as much as [24]. Anyway,
Das [26] estimated that IoT semiconductor
manufacturing alone used 2 EJ in 2016 and will use
35 EJ in 2025. This means rising from around 0.3%
to 5% of GPEC. Using Table 2, the related CO2
emissions would be ≈0.12 Gt (2 EJ×0.06 Gt/EJ) in
2016 and ≈1.9 Gt (35 EJ×0.055 Gt/EJ) in 2025,
respectively. Such hypotheses are much in line with
those trends presented in [24]. The current
hypothesis for 2020 based on updates of [16] is that
manufacturing will use around 300±50 TWh
emitting 0.17±0.04 Gt CO2.This is 0.83 Gt less than
[24]indicating that production electricity CO2and
otherCO2 are much underestimated in [16]. In 2030
this is assumed to decline to 240±60 TWh emitting
0.13±0.03 Gt CO2.However, the uncertainty is
hugeif the findings in [24] and [26] are considered.
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
54
Volume 15, 2020

If including the production related “truncation error”
from [24], the numbers for 2020 are 0.99±0.06 Gt
and 0.95±0.05 Gt. If including the hypotheses and
findings of Das[26], the “truncation error” for CO2
emissions would be even higher.
6 Results
The main results are obtained by combining the CO2
intensities in Table 5 with the TWh in Sections 5.1-
5.5 for each sector for 2020 and 2030. SimaPro
version 9.0.0.31 is used to obtain the uncertainty
ranges. Table 6 shows the summary of digital sector
CO2 evolution and the share of the digital Sector of
global CO2 emissions.
Table 6: Approximate million tonnes CO2 evolution
from digital sectors
Digital Sector Year 2020 Year 2030
0%
renewables
Data Centers
use
160±25 420±120
Mobile
networks use
54±13 170±60
Optical
networks use
83±20 150±60
Devices use 460±110 410±150
Manufacturing
processes
1000±60 960±50
TOTAL (Gt)
0%
renewable
power for
Networks and
Data Centers
in 2020
1.76±0.17 2.11±0.3
TOTAL (Gt)
5%
renewable
power for
Networks and
Data Centers
in 2030
2.08±0.3
TOTAL (Gt) 1.76±0.35
50%
renewable
power for
Networks and
Data Centers
in 2030
Share of
Global CO2
emissions,
0%
renewables in
2020
4.7% 5.2%
Share of
Global CO2
emissions,
5%
renewables in
2030
5.2%
Share of
Global CO2
emissions,
50%
renewables in
2030
4.4%
Monte Carlo simulations in SimaPro version
9.0.0.31 - each with 100000 runs - show 3.89%,
4.28% and 18.1% probabilities that the CO2will
emissions decrease between 2020 and 2030 in the
0%, 5% and 50% renewables scenarios,
respectively. This means that the probability is
82%to 96% that CO2 emissions from the digital
sector will increase between 2020 and 2030.Fig. 4
shows a graph for the 0% renewable scenario
resulting from the present simulation.
Fig. 4. Monte Carlo simulation result for probability
of digital sector CO2 reduction between 2020 and
2030.
0,
0,01
0,02
0,03
0,04
0,05
0,06
‐0,618
‐0,5844
‐0,5508
‐0,5172
‐0,4836
‐0,45
‐0,4164
‐0,3828
‐0,3492
‐0,3156
‐0,282
‐0,2484
‐0,2148
‐0,1812
‐0,1476
‐0,114
‐0,0804
‐0,0468
‐0,0132
0,0204
0,054
0,0876
0,1212
0,1548
Probability
'DigitalSectorFootprint2020'minus
'DigitalSectorFootprint2030'
Confidenceinterval:95%
Resultof100000runMonteCarloSimulationinSimaPro
9.0.0.31for0%renewablesin2030
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
55
Volume 15, 2020

7 Discussion
Several authors in the fundamental semiconductor
research area see problems and potential solutions to
the increasing power use in computing.
7.1 Drivers and uncouplers
Obviously, Moore’s law does not hold
anymore[27].Still,Das argued that IoT
semiconductor devices use stage power will sharply
decline until 2025 [26]. Thylen [28] argued that
nanophotonics is one of the solutions to the
emerging energy per bit problem in data centers and
optical networks. Further, material breakthroughs
are necessary to change nanophotonics [28].
Moreover,it is very reasonable to assume that there
will be several engineering tricks such as replacing
 electronic (de)serializers time division
multiplexers with optical space division
multiplexing.
 clock and data recovery by synchronizing
photonic pulses[28]
However, much of these “tricks”may already have
beenimplemented?Directly modulated low-cost
vertical cavity surface emitting lasers (VCSELs) -
used as optic links in datacom- use 56-510
fJ/bit[29]. If such efficiencies are comparable to
“what it takes” on system level (≈50 fJ/operation
[5]) it is not all unreasonable that computing will
have a reasonable power consumption in this
decade. The discussion about implications of current
switching energy (J/transistor) roadmaps is hugely
important. It has been predicted[5] that even with
PUE=1 that we are in need of a new switch or chip
architecture.That is, predictions of electricity use per
computation for the next decade is key as well as the
implications of such predictions when combined
with number of computations.
The projected global instructions per second (in
CPUs and GPUs) – in 2030 is around 2.6 Yotta
(2.6×1024) instructions per second[6] andthe
instructions per Joule (Instructions per kWh[7]) in
2030 imply thousands of TWh (as shown in Table 1
and Fig.1)for computing in 2030[5]. The often cited
formula from Koomey about computing energy
efficiency[7], predicting Instructions/kWh may still
hold validity[5], but it has not been confirmed
lately.Industry is projecting that more and more W
is needed per server to reach the required
performance. Moreover, the FLOP/s/W for servers
and computers are decreasing. Likewise are the
operations per second per Watt decreasing.
Operations per second per Wattis often used for data
center energy performance [5].More concerningly,
the maximum heat fluxes (W/cm2) in commodity
CPU / GPUs have been reached, the clock
frequencies reached plateus in the 2010s, and the
Voltage reduction is slowing [30]. The relation
between the manufacturing CO2 and the use stage
CO2 for the whole digital sector has been assumed
to be on average 25% manufacturing and 75%
use[16]. Using 1 Gt CO2per year for
Manufacturing[24]wouldeither render 4 Gt CO2per
year in total for the digital sector or a 50% share for
Manufacturing. Anyway 1 GtCO2per year would
add around 0.8 Gt CO2per year, showing a
tremendous underestimation of the sector in 2015.
Such Gt CO2per year would be close to worst case
scenario of [16] for Production in 2020.
Theremarkable hypotheses [26] of 35 EJ for IoT
semiconductor manufacturing in 2025 also indicate
that manufacturing/production could be largely
underestimated in these kinds of CO2 predictions.
Table 7 shows some other TWh results for networks
and data centers obtained from updating estimations
for 2015 and 2020 [16] and using average of values
from [5] for 2025 and 2030. These results are
further developed in this article.
Table 7. Approximate TWh evolution from selected
ICT Sectors
2015 2020 2025 2030
Mobile networks
use stage 152 136 286 700
Optical networks
use stage 179 171 138 146
Data centers use
stage 220 207 469 799
From Table 7 it could already be argued that the
CO2 emissions for digital networks will not be
reduced easily in the short-term if some global
average energy mix is used.
Also there is a need for a better understanding of
how everyday online practices are shifting[13].
Take self-driving cars in which 4 persons – which
may include the “driver” - watch streamed 8K 4D
videos hour after hour. Such business cases may add
to the growing video traffic. Also epidemics and
pandemics may lead to more video conferences and
leisure streaming as a result of quarantine situations.
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
56
Volume 15, 2020

7.2 Renewables?
Buying renewables “somewhere else” leads to more
renewables in the average macro grid mix which
reduces CO2 emissions in the long-run.
Hypothetically CO2 reduction in the short-term can
be achieved by local renewable power production
which feeds directly the needs of own of facilities
and equipment.
As shown in Section 6, 50% local and direct
renewable electricity supply to all global data
centers and networks in 2030 does not likely (18%
chance) reduce the digital sectors own, relatively
small, CO2 emissions between 2020 and 2030.
7.3 Steel production with hydrogen – 10% of
GEU in 2030?
In this section the additional electricity use and CO2
emissions of hydrogen production for steel
production is estimated. The motivation is that Steel
is an important material constituent also for
computing equipment and the need to highlight
potential new large electricity demands. In order to
reduce the CO2emissions from steel production,
≈3.3 Gt[31], it has been proposed to produce steel
with hydrogen [32][33] instead of coke [34].
How much additional electricity would be required
globally – and CO2 emitted - if all global steel
would be produced with hydrogen reduction (2)
instead of coke reduction (1)?
2 Fe2O3 + 3 C  4 Fe + 3 CO2 (1)
Fe2O3 + 3 H2 2 Fe + 3 H2O (2)
In 2018, the total world crude steel production was
1808.6 million tonnes (Mt) and Sweden’s steel
production was4.7 Mt [35].
The emissions from steel production is 1.1 Gt
CO2per year according to (1): 1.8×1012 kg Fe×0.59
kg CO2/kg Fe. This shows that (1) is very crude and
underestimatethe actual CO2 emissions[31].
Anyway, according to (2),producing 1 kg Fe
requires ≈0.053 kg H2. It takes around 36 kWh
electricity to produce 1 kg H2[36]. Accordingly,
≈1.93 kWh/kg Fe, and 3492 TWh (1.8×1012 kg
Fe×1.93 kWh/kg), i.e. almost 10% of GEU in 2030.
3492 TWh global annual electricity use derived
from (2) would according to Table 5 emit ≈1.85 Gt
CO2 per year. This represent a 68% increase
compared to (1) but a 44% decrease compared to
[31]. In Sweden, 9 extra TWh (≈5% of Sweden
electricity use [3])– i.e. corresponding to one more
1000 MW nuclear reactor –will be required.
Moreover, the global steel production may grow
from 2018 to 2030. The CO2 balance - using (1)
and (2), effect of recycling, etc. - of current and
future steel production is beyond the scope of this
article. Also the effect on the digital sector and
computing footprint of hydrogen production is
excluded. The section suggests thatthe hydrogen
economy will require much extra electricity.
However, using hydrogen for steel production could
make sense from a CO2 emission viewpoint.
8 Conclusion
The probability is between 82 and 96% that
computing and the digital sector will increase its
CO2 emissions between 2020 and 2030. However, at
4-5%, computing and digital emissions will remain
a relatively small share of the total global in this
decade.
9 Next steps
Obviously there are several assumptions which
should be revisited. For instance, it is not self-
evident that the overall power used by devices in
homes will decline in this decade. The potential
effect of waning Moore’s law for the use stage
ofsuch consumer devices is only included broadly in
[5]. However, this effect was neglected by
frameworks such as those used by [16] and [25].The
framework used by [21] for end-user devices may
include this effect. More global estimations of
instructions and operations – and measurements of
J/operation - are required to substantially move this
field forward.The degree to which home-owners
will be able to produce own renewable power,
which can run their digital devices, could also be
investigated.
References:
[1] U.S. Energy Information Administration. EIA
projects 28% increase in world energy use by
2040. [cited 2020 March 4]: Available from:
https://www.eia.gov/todayinenergy/detail.php?i
d=32912
[2] U.S. Energy Information Administration. EIA
projects global energy-related CO2 emissions
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
57
Volume 15, 2020

will increase through 2050. [cited 2020 March
3]: Available from:
https://www.eia.gov/todayinenergy/detail.php?i
d=41493
[3] L. Bengtsson. Vad händer med klimatet? (In
Swedish), Karneval Förlag, 2019.
[4] BP Statistical Review of World Energy. June
2019. 68th Edition. [cited 2020 March 4]:
Available from:
https://www.bp.com/en/global/corporate/energ
y-economics/statistical-review-of-world-
energy.html
[5] A.S.G. Andrae. 2019. Prediction studies of the
electricity use of global computing in 2030.
International Journal of Science and
Engineering Investigations, Vol.8,No.86, pp.
27-33. http://www.ijsei.com/papers/ijsei-
88619-04.pdf
[6] Z.W. Xu. 2014. Cloud-sea computing systems:
Towards thousand-fold improvement in
performance per watt for the coming zettabyte
era. Journal of Computer Science and
Technology, Vol.29,No.2, 177181.
[7] J. Koomey, S. Berard, M. Sanchez, H. Wong.
2011. Implications of historical trends in the
electrical efficiency of computing, IEEE
Annals of the History of Computing,
Vol.33,No.3, pp. 46-54.
[8] D.A. Miller. 2017. Attojoule optoelectronics
for low-energy information processing and
communications. Journal of Lightwave
Technology, Vol.35,No.3, pp. 346-396.
[9] C. Ekberg, D.R. Costa, M. Hedberg, M.
Jolkkonen.2018. Nitride fuel for Gen IV
nuclear power systems. Journal of
radioanalytical and nuclear chemistry, Vol.
318,No.3, pp. 1713-1725.
[10] OECD/NEA.2014. GEN IV International
Forum. Technology roadmap update for
generation IV nuclear energy
systems.https://www.gen-
4.org/gif/upload/docs/application/pdf/2014-
03/gif-tru2014.pdf
[11] Q. Li, Y. Liu, S. Guo, H. Zhou. 2017. Solar
energy storage in the rechargeable
batteries. Nano Today, Vol.16, pp. 46-60.
[12] M. Pärssinen, M. Kotila, R. Cuevas, A.
Phansalkar, J. Manner. 2018. Environmental
impact assessment of online
advertising. Environmental Impact Assessment
Review, Vol. 73, pp. 177-200.
[13] J. Morley, K. Widdicks, M. Hazas. 2018.
Digitalisation, energy and data demand: The
impact of Internet traffic on overall and peak
electricity consumption. Energy Research &
Social Science, Vol.38, pp. 128-137
[14] GeSI. 2019. Digital with a Purpose – delivering
a smarter 2030. [cited 2020 March 4]:
Available from:
http://digitalwithapurpose.gesi.org/platforms/di
gital-with-a-purpose-delivering-a-smarter2030
[15] Consumer Technology Association. 2019. 2019
Industry Report on GHG Emissions.[cited 2020
March 4]: Available from:
https://cdn.cta.tech/cta/media/media/resources/
cta_ghg_report.pdf.
[16] A.S.G. Andrae, T. Edler.2015.On global
electricity usage of communication technology:
trends to 2030. Challenges, Vol.6,No.1, pp.
117-157.
[17] BP Statistical Review of World Energy. June
2017. 66th Edition. [cited 2020 March 4]:
Available from:
https://www.bp.com/en/global/corporate/energ
y-economics/statistical-review-of-world-
energy.html
[18] Reuters. China's internet data power usage to
surge through 2023: study. 2019.[cited 2020
March 4]: Available from:
https://www.reuters.com/article/us-china-
carbon-internet/chinas-internet-data-power-
usage-to-surge-through-2023-study-
idUSKCN1VU06A
[19] L. Zhang, J. Chen, E. Agrell, R. Lin, L.
Wosinska. 2020. Enabling Technologies for
Optical Data Center Networks: Spatial Division
Multiplexing. Journal of Lightwave
Technology, Vol.38.No.1, pp. 18-30.
[20] R.B. Mitchell, R. York. 2020. Reducing the
web's carbon footprint: Does improved
electrical efficiency reduce webserver
electricity use?, Energy Research & Social
Science, Vol.65, 101474.
[21] L. Stobbe, N.F. Nissen,J. Druschke et al.
2019.Methodology for Modeling the Energy
and Material Footprint of Future
Telecommunication Networks, Going
GreenEcoDesign, Yokohama, Japan, Nov. 25-
27.
[22] A.S.G. Andrae.2019. Comparison of Several
Simplistic High-Level Approaches for
Estimating the Global Energy and Electricity
Use of ICT Networks and Data Centers.
International Journal of Green Technology,
Vol.5,No.1, pp. 51-66.
[23] ADVA Sustainability Report 2018. p.46. cited
2020 March 4]: Available from:
https://www.adva.com/-/media/adva-main-
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
58
Volume 15, 2020

site/resources/sustainability/sustainability/pdfs/
sustainability-report-2018-english.ashx
[24] L. Cabernard. 2019.Global supply chain
analysis of material-related impacts in ICT
(MRIO approach). [cited 2020 March 3]:
Available from:
http://www.lcaforum.ch/portals/0/df73/DF73-
04_Cabernard.pdf
[25] L. Belkhir, A. Elmeligi. 2018. Assessing ICT
global emissions footprint: Trends to 2040 &
recommendations. Journal of Cleaner
Production, Vol.177, pp. 448-463.
[26] S. Das. 2019. Global Energy Footprint of IoT
Semiconductors. [cited 2020 March 4]:
Available from:
http://www.lcaforum.ch/portals/0/df73/DF73-
09_Das.pdf
[27] C.M. Schneider. 2020. Spintronics: Surface and
Interface Aspects. Surface and Interface
Science: Volume 9: Applications I/Volume 10:
Applications II
[28] L.G. Thylen. Integrated Nanophotonics, the
Quest for Novel Photonics and Electronics
Materials, Monolithic Electronic/Photonic
Integration and Applications in Power Hungry
Data Centers. [cited 2020 March 3]: Available
from: https://inphyni.cnrs.fr/en/news-and-
events/seminars/seminaire-de-lars-
thylen/@@highlight_view#.XlvR-qhKg2w
[29] K. Szczerba, P. Westbergh, J.S. Gustavsson, M.
Karlsson, P.A. Andrekson, A. Larsson. 2015.
Energy efficiency of VCSELs in the context of
short-range optical links. IEEE Photonics
Technology Letters, Vol.27,No.16, pp. 1749-
1752.
[30] R. Waser, Nanoelectronics and information
technology, Wiley-VCH Verlag GmbH, 2003.
[31] B.J. Van Ruijven, D.P. Van Vuuren, W.
Boskaljon, M.L. Neelis, D. Saygin, M.K. Patel.
2016. Long-term model-based projections of
energy use and CO2 emissions from the global
steel and cement industries. Resources,
Conservation and Recycling, Vol.112, pp. 15-
36.
[32] D. Kushnir, T. Hansen, V. Vogl, M.
Åhman2020. Adopting hydrogen direct
reduction for the Swedish steel industry: A
technological innovation system (TIS) study.
Journal of Cleaner Production, Vol.242,
118185.
[33] E. Karakaya, C. Nuur, L. Assbring. 2018.
Potential transitions in the iron and steel
industry in Sweden: Towards a hydrogen-based
future?, Journal of Cleaner Production,
Vol.195, pp. 651-663.
[34] P.A. Renzulli, B. Notarnicola, G. Tassielli et al.
2016. Life cycle assessment of steel produced
in an Italian integrated steel mill.
Sustainability, Vol.8,p. 719.
[35] Wikipedia. 2020. List of countries by steel
production.
https://en.wikipedia.org/wiki/List_of_countries
_by_steel_production.
[36] K. Zeng, D. Zhang. 2010. Recent progress in
alkaline water electrolysis for hydrogen
production and applications. Progress in
Energy and Combustion Science, Vol.36,No.3,
pp. 307-326.
WSEAS TRANSACTIONS on POWER SYSTEMS
DOI: 10.37394/232016.2020.15.6
Anders S. G. Andrae
E-ISSN: 2224-350X
59
Volume 15, 2020