AI-Driven Data Centers: Revolutionizing

Decarbonization Strategies

2

Contents

About the Authors 3

Global Impact 4

AI-Driven Data Centers: Revolutionizing Decarbonization Strategies

Chapter 1: Energy Sector 11

Data Centers are Essential to AI-driven Decarbonization of Power Sector

Chapter 2: Manufacturing Sector 13

Data Centers are Essential to AI-driven Decarbonization of Production and Manufacturing

Chapter 3 : Transportation Sector 15

Data Centers are Essential to AI-driven Decarbonization of Transport

Chapter 4: Buildings Sector 17

Data Centers are Essential to AI-driven Decarbonization of Buildings

Resources 18

3

Marcin Wegrzyn With more than 10 years of experience R&D experience in academia and industry, Marcin Wegrzyn is a Research Analyst for the Energy Management Research Center at Schneider Electric. He focuses his considerable expertise primarily on sustainability in data centers, data centers as a whole, and on commercial and residential buildings. He holds a Master of Polymer Technology, a Master of Business, and a PhD in Industrial Engineering and Production.

Steven Carlini

Steven Carlini is the Vice President of Innovation and Data Center at Schneider Electric. With extensive global experience, Steve leads the Energy Management’s Office of Innovation and Data Center Solutions, a team focused on spearheading Schneider’s data center, digital energy, and residential businesses. In 2023, Steve was named to Capacity Media's Capacity POWER 100, a list of the “trailblazers, innovators and leaders driving the global digital infrastructure space.” In 2022, he joined the Forbes Technology Council. In 2020, he was named to Data Economy’s Future 100, the top 100 people to watch in the next decade. He is a member of the World Economic Forum’s 5G-Next Generation Networks Programme. Results- oriented, Steven transforms ideas into products, solutions and systems. His areas of focus include innovation, AI, hydrogen, sustainability, 5G and 6G, cloud and edge computing, DCIM, BMS, and EDMS

About the Authors

4

Mitigating global warming and reducing carbon emissions are important goals for industries. IT and advanced software housed in data centers play a critical role in decarbonization in nearly every sector of the economy. The continuously evolving capacity for computational power in centralized or edge data centers can be used to analyze emissions and make tailored recommendations, ultimately leading to carbon footprint reductions. Automating energy-intensive processes in the manufacturing industry, whether it is to improve overall efficiency or coordinate logistics and supply chains to reduce emissions from transportation, is available thanks to implementation of algorithms of Artificial Intelligence (AI) hosted in data centers. Also, the evolution of chips and other components of IT infrastructure enables more FLOPS (floating-point operations per second) for more powerful and effective AI algorithms for process improvements and efficiencies.

The data center sector is increasingly gaining energy efficiency, and this trend is likely to continue. Strategic deployment of colocation data centers and the building of an optimal network of nodes can solve data transfer and latency issues. Making the IT infrastructure accessible and reliable improves efficiency in other sectors of the economy that use data centers. They can work on low carbon energy sources while providing broader access to predictive analytics and repair automation to streamline maintenance and services with reduced emissions. Skeptics of data centers list key challenges like emissions from energy sources and inefficient cooling along the acoustic and hardware waste or negative impact on biodiversity.1 Data centers are rapidly evolving, and these accusations are countered with self-driven or regulation- induced developments that have already guided performance towards a zero-carbon scenario. The fast growth of data centers as a response to digital transformation and an initially limited number of regulations in the sector caused server utilization rates far under capacity.2

Many groundbreaking technologies were not accepted by society at an early stage of their implementation. Data centers may share a similar history. Many people initially feared electricity, accusing it of being dangerous. Technology evolution and innovations, like developing a method to generate alternating current or moving high voltage wires in cities from overhead to underground, have increased safety and resiliency of electricity. Data centers are at pivotal point, proving their significance in advancing human civilization.

Even though many factors are involved in estimations, scenarios can change dramatically before 2050 and leveraging computational power to help solve some of the world’s most urgent environmental challenges will remain important. It requires the presence of data centers in much greater number than today’s count. This perspective is part of a series addressing data center-driven decarbonization of top polluters in today’s business.

Electrification

Substituting technologies based on fossil fuels with solutions powered by electricity. Key benefits of electrification are higher process efficiency, reduction of greenhouse gas emissions, and straightforward digitalization

Global Impact

1Network King, Five areas of environmental impact in data centers (2022)

2IBM, Are Your Data Centers Keeping You from Sustainability? (2022)

https://network-king.net/five-areas-of-environmental-impact-in-data-centres/ https://www.ibm.com/blog/are-your-data-centers-keeping-you-from-sustainability/

5

Examples of electrification:

• replacing combustion engine vehicles with electric vehicles, • replacing oil heaters with heat pumps in residential sector, • introducing induction furnaces for industrial processes as an alternative to blast

furnaces.

Data Centers For Sustainability

Greenhouse gasses (GHG) emissions must be reduced by 50% by 2030 and zeroed by 2050.3 Electrification is the foundational solution for many applications because it enables assets to be directly powered by renewable resources, leaving a greatly reduced carbon footprint in the future. However, this decarbonization requires digitization of industries, processes, and applications.

Forecast of data centers in the modern economy

The enablement effect of the IT sector on emissions will facilitate carbon footprint reduction of the highest emitting industries.4 Reduction is amplified by the digitization of transport, industry, and buildings, and a new generation of consumers potentially making more sustainable choices. Dedicated technology developments like AI, digital twin, hybrid cloud, and wider introduction of 5G network opening possibility more sophisticated operations will secure the current trend in data center evolution.5 Secure and complex blockchain operations, software development and operations (DevOps) activities, and computing processing of power-heavy AI algorithms require sufficient computational power and fast, reliable data transfer.

Evolution of the IT sector and data centers may follow a range of optimistic or pessimistic scenarios. As expected, the pessimistic view predicts an overall energy consumption increase across economic sectors, driven by changing consumer behaviors and the growth of increasingly digital business models.6 The Jevons Paradox7 indicates this activity increases carbon footprint given a pessimistic view; where an increase in data centers and a concomitant increase in their energy requirements improve the economy but damage the environment. Yet, this is becoming increasingly unlikely, as data centers are doing the work to improve their operational efficiencies and energy mix.4

3 Berkley Lab, A 50% Reduction in Emissions by 2030 Can be Achieved (2022) 4 This is described in detail in a series of Schneider Electric White Papers on decarbonizing: power sector (White Paper 180), manufacturing (White Paper 167), transportation (White Paper 168), buildings (White Paper 169). 5 Uptime Institute, Five Data Center Predictions for 2023 (2022) 6 Universität Munster, The Center of Interdisciplinary Sustainability Research (2023) 7 Patterns, The real climate and transformative impact of ICT: A critique of estimates, trends, and regulations (2021)

https://newscenter.lbl.gov/2022/06/02/emissions-reduction-roadmap/ https://uptimeinstitute.com/resources/research-and-reports/five-data-center-predictions-for-2023 https://www.uni-muenster.de/Nachhaltigkeit/en/ https://www.sciencedirect.com/science/article/pii/S2666389921001884 https://www.sciencedirect.com/science/article/pii/S2666389921001884

6

Tremendous power consumption without increase in emissions

Figure 1 Forecast of data center power consumption and emissions. (Schneider Electric estimates)

Scenarios where data centers show a positive environmental impact are becoming more likely - given these same efficiencies. When sectors employ AI to understand, manage and reduce their carbon footprint from operations, the efficiency gain is higher than expected in many cases. As sectors pursue and create AI-enabled solutions through their data centers, they support cross the portfolio optimizations - where environmental impacts are reduced twice over. Once in the data center and again on deployment in industry. These approaches support the upside scenario - where data centers deliver a more efficiency in the organization and in its energy mix.

Global data continues to rise driven by AI and digital lifestyles. Large amounts of information will be transferred to and processed in data centers, which will have to continue growing in number and capacity. Skepticism on the negative impact of that development on the environment can be mediated with information in Figure 1. Our projection referring to data center sector, from smallest local edge to large enterprise data centers, is based on different scenarios forecasting low emissions intensity from electricity production in 2050 (Electricity 4.0), close to 0.1 kg CO2/ kWh. 8,9 Centralized and edge data centers produce around 330 MtCO2/year,10 but even with continuous build-out, we project that carbon emissions will track around that number for the next 30 years due to efficiency improvements and ramping up of renewable electricity supply (see Figure 2). A slight disruption of data center emissions and power demand curves is caused by a similar trend in data processed forecast, between 2020 and 2030.11 Increased computing intensity thanks to power and computation-intensive training of AI algorithms causes a single, short-term impact.

Electricity 4.0

Combination of electricity potential, a superior to fossil fuels power source, and digital tools allowing smart management of infrastructure.

Electricity is proven to be up to 5x more efficient than other energy sources thanks to fewer losses along the supply network. Electricity is also the best vector for decarbonization that opens the possibility to form a network of devices connected to Internet. AI-aided smart management of supply and demand provides process efficiency and relieves main power grid.

8 Nature Communications, A global comparison of building decarbonization scenarios by 2050 towards 1.5–2 °C targets (2022) 9 Ener data, CO2 intensity of electricity generation (2023) 10 IEA, Data Centres and Data Transmission Networks (2023) 11 Schneider Electric, Tradeoff Tool: Data Center & Edge Global Energy Forecast (2021)

https://www.nature.com/articles/s41467-022-29890-5 https://www.nature.com/articles/s41467-022-29890-5 https://eneroutlook.enerdata.net/forecast-world-co2-intensity-of-electricity-generation.html https://www.iea.org/energy-system/buildings/data-centres-and-data-transmission-networks https://www.se.com/ww/en/work/solutions/system/s1/data-center-and-network-systems/trade-off-tools/data-center-and-edge-global-energy-forecast/

7

Adding to global decarbonization will be a shift to renewable power sources.

Figure 2 Change of energy sources to renewables – optimistic and pessimistic scenarios. Schneider Electric, Back to 2050 (2021)

Experts estimate that data processing, storage, and transmission use 1% of global electricity.8 This share has hardly changed since 2010, even though the number of internet users has doubled, and global internet traffic has increased 15-fold, according to the International Energy Agency.12 Unchanged emissions from data centers and growth of the IT sector with rapid increase of processed data in the coming decades in fact decreases the unitary net carbon footprint of data centers. Required for cloud, enterprise, telecom, and emerging technologies, data centers need another level of reliability, quality, and physical infrastructure design to reach climate change objectives. Lowering overall consumption of energy is a challenge for data center operators. This requirement exists along with lower carbon intensity of the infrastructure.

Components of AI, including the recent rise of generative AI, impact not only data center energy use, but also the related pathway of infrastructure development. New AI- related trends drive new paradigm of the data center evolution concerning AI training and AI implementation. This process is described in greater detail in Schneider Electric White Paper 110, “The AI Disruption: Challenges and Guidance for Data Center Design”.

The IT sector is tightly bonded with production and other energy-intensive branches of the economy. Solutions tailored for sector-specific problems often show the unprecedented position that data centers have in modern economy.

Data centers support decarbonization of the economy

Decarbonization of economy will most likely be an outcome of assets electrification and digitalization.

Important aspects that need to be addressed in a discussion on the effect of data centers on sustainability are the biggest polluters with their key challenges and top drivers for economy decarbonization. The biggest polluter is the energy sector, with 15.8 GtGHG emissions in 2022, followed by transport (8.4 GtGHG) and manufacturing

12 IEA, Data centers and energy – from global headlines to local headaches? (2019)

https://www.se.com/ww/en/download/document/SPD_WP110_EN/ https://www.se.com/ww/en/download/document/SPD_WP110_EN/ https://www.iea.org/commentaries/data-centres-and-energy-from-global-headlines-to-local-headaches

8

(6.3 GtGHG).12 Decarbonization must start with electrification of processes and operations to improve energy efficiency and logistics, focusing on the sector- dedicated regulatory environment. A green future will come with digitalization and a wider application of AI utilizing sensors within Internet of Things (IoT) along fast data transfers. Only electrification will allow reaching decarbonization targets and follow the optimistic scenario of sustainable economy. It is inevitable that decarbonization requires digitization, and this will not happen without broad implementation of data centers.

Decarbonization begins with a dramatic shift from fossil fuels power sources to electricity by 2050.

Figure 3 Change of global energy sources – optimistic and pessimistic scenarios. Schneider Electric, Back to 2050 (2021)

We predict a 28 Gt reduction of global GHG emissions from all industries by 2050. The role that data centers play in decarbonization is fundamental to meet the optimistic scenario. Even with a skeptical and conservative estimate that 10% of decarbonization is dependent on data centers, the benefit is 2800 Mt removed from economy for 367 Mt emitted. An optimistic estimate of 30% shows a benefit of 8400 Mt for 367 Mt emitted. Data centers compose negligible share of total global emissions, and this value is predicted to remain nearly unchanged in the coming decades. The expected growth of electricity use in most sectors thanks to transformation of energy generation and distribution (see Figure 3, Figure 4) require implementation of technologies hosted in data centers.13 Emissions are not aligned with energy demand. Industry growth, enabled by AI algorithms housed in data centers, will cause larger participation in total emissions when following pessimistic scenario. Analogous situation is expected for buildings and transport, where digital technologies housed in data centers participate in reducing carbon footprint, allowing to cumulatively reduce 84% global emissions in 2050.

13 Schneider Electric, Back to 2050 (2021)

https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp

9

As we see buildings, transportation and industrial application electrify, the optimistic scenario forecasts reduction in power demands

Figure 4 Change in energy demand for different industries – optimistic and pessimistic scenarios. Schneider Electric, Back to 2050 (2021)

Artificial Intelligence for Sustainability

From machine learning representing narrow AI (ANI) to the so-far-unreachable machine consciousness strong AI (ASI), Artificial Intelligence has opened new possibilities in various areas. Deep learning extracting and clarifying components of data is used to decarbonize the energy grid in NAM,14 robotics impersonating human actions participate in decarbonizing cities in Asia15 and oceans,16 while random forest algorithms using decision trees provide energy stability in developing countries and help with electrical vehicles charging load predictions.17 Simple machine learning provides technology to reduce emissions from the energy grid18 and from the shipping industry.19 Expert systems imitating human decisions improve operations in Japan’s steel industry.20 Fuzzy logic validating decisions aid decarbonization of urban areas in Europe21 and support assessment of corporate sustainability in the manufacturing sector.22 These are all branches of AI already heavily used in the sustainability journey. It is only the beginning. As more businesses join, more impact will be reported, providing more sustainable data centers are available to handle necessary tasks.

14 Energies, A Deep Learning Approach for Exploring the Design Space for the Decarbonization of the Canadian Electricity System (2023) 15 Journal of Environmental Management, Towards low-carbon development: The role of industrial robots in decarbonization in Chinese cities (2023) 16 Marne Pollution Bulletin, Ocean oil spill detection from SAR images based on multi-channel deep learning semantic segmentation (2023) 17 Energies, The Application of Improved Random Forest Algorithm on the Prediction of Electric Vehicle Charging Load (2018) 18 Sustainability, Machine Learning Techniques for Decarbonizing and Managing Renewable Energy Grids (2022) 19 Procedia Computer Science, A Comparative Research of Machine Learning Impact to Future of Maritime Transportation (2019) 20 IFAC Proceedings, A Hybrid Expert System Combined with a Mathematical Model for BOF Process Control (1992) 21 Energy Conversion and Management, A new fuzzy model of multi-criteria decision support based on Bayesian networks for the urban areas' decarbonization planning (2022) 22 Sustainability, A Fuzzy Logic-Based Tool for the Assessment of Corporate Sustainability: A Case Study in the Food Machinery Industry (2017)

https://www.mdpi.com/1996-1073/16/3/1352 https://www.mdpi.com/1996-1073/16/3/1352 https://www.sciencedirect.com/science/article/abs/pii/S030147972300004X https://www.sciencedirect.com/science/article/abs/pii/S030147972300004X https://www.sciencedirect.com/science/article/abs/pii/S0025326X23000826 https://www.sciencedirect.com/science/article/abs/pii/S0025326X23000826 https://www.mdpi.com/1996-1073/11/11/3207 https://www.mdpi.com/1996-1073/11/11/3207 https://www.mdpi.com/2071-1050/14/21/13939 https://www.sciencedirect.com/science/article/pii/S1877050919312128 https://www.sciencedirect.com/science/article/pii/S1877050919312128 https://www.sciencedirect.com/science/article/pii/B9780080417042500170 https://www.sciencedirect.com/science/article/pii/B9780080417042500170 https://www.sciencedirect.com/science/article/pii/S0196890422008251 https://www.sciencedirect.com/science/article/pii/S0196890422008251 https://www.mdpi.com/2071-1050/9/4/583 https://www.mdpi.com/2071-1050/9/4/583

10

Cloud to edge computing enable the use of sophisticated tools supporting carbon footprint reduction in every area of human activity, i.e., transport, industry, or power generation. Independently, if based on a centralized network and large data centers, or on a distributed network of edge nodes like cars, smartphones, or smart home devices, AI plays a key role in driving individual users’ actions towards a net-zero emissions goal.

The evolution of power sources and distribution will not happen without the presence of the IT sector. Electrification without digitalization supporting power management and data control is unrealistic. Considering large units as well as local edge data centers, AI algorithms need a reliable and secure infrastructure. Edge AI, an alternative to on- premises AI and cloud AI, is expected to play a vital role in economy transformation with real-time analyses of data from localized sensors. Gartner claims that by 2025, as much as 75% of data will be generated outside of centralized facilities.23 Cloud AI offers scalability and carries a potential to process copious amounts of data required for model accuracy. Some bespoke benefits from using edge AI include reduced cost and improved performance of business activities, real-time analytics, easier or enabled product servitization, facilitated IT/OT integration, and improvement of machine learning algorithms. Security risk and the inaccuracy of some models along with limited processing power and heterogeneous characteristics of edge environments still need to be resolved before broader adoption of the edge AI technology.

Recently observed growth of AI models generally leads towards a clear goal – induction of behavioral evolution, making Green AI a new standard. Algorithms that are environmentally friendly will solve problems with necessary access to large amounts of data. This can be challenging because codes are not yet optimized in embedded systems or constrained by limited resources.24

Green AI can be achieved with several scenarios. Strategic selection of data center location, where machine training takes place, may have a strong impact on the overall carbon footprint of the project. This strategy will benefit mostly time- and latency- insensitive, large workloads. Solution for simpler algorithms is in reduction of the amount of machine learning models through quantization and knowledge distillation, or in using models partitioned into discrete regions with fewer features.

Conclusion

Data centers are cornerstones of digital economy. By improving efficiency and sustainability parameters, data centers become a low GHG emitting solution to reduce carbon footprint in top polluting sectors of human activity like transportation, manufacturing, and power generation. Predicted further evolution of data centers will result with a more efficient ratio between energy demand and emis-sions, providing green computational power instrumental for decarbonization. Adaptation of smart algorithms forecasted to decrease global carbon footprint by orders of magnitude require data centers in AI training and inference. The most advantageous impact is realized when optimization occurs across different electrified industries, making AI an essential component that helps to address the complexity of this process. There is no chance to follow the optimistic scenario reaching desired levels of sustainability and ecology in human activity without broader implementation of data centers.

Read White Paper 106: "AI-Driven Data Centers Revolutionizing Decarbonization Strategies.

23 Gartner, Innovation Insight for Edge AI (2022) 24 PACIS 2023 Proceedings, Organizational Adoption of Green Artificial Intelligence: An Institutional Perspective (2023)

https://www.se.com/ww/en/download/document/SPD_WP106_EN/ https://www.gartner.com/smarterwithgartner/innovate-with-edge-ai https://aisel.aisnet.org/cgi/viewcontent.cgi?article=1141&context=pacis2023 https://aisel.aisnet.org/cgi/viewcontent.cgi?article=1141&context=pacis2023

11

Energy sector is the largest polluter in the economy responsible for 35% of today’s emissions.25 Regulatory pressure from governments drives decarbonization of energy generation and distribution, causing an evolution towards clean energy. A direct pathway to reducing sectorial carbon emissions focuses on a widespread transition to renewables. This move can provide only partial decarbonization and the rest must go through digitalization of power sector. Data centers are the unsung heroes in the evolution towards a net-zero economy and green power. Artificial Intelligence (AI) proven to increase process efficiency26 is forecasted to drive decarbonization of the power sector by coordinating power delivered in a distributed grid – a process that is becoming more complex with extensive decentralization. Generative AI, which is offering more advanced and intuitive models allowing unsupervised predictions in decarbonization journey, requires higher workloads and substantial computational resources. Resulting data generation and processing will induce much broader deployment of IT infrastructure: local edge and centralized data centers.

Up to 57% increase of global power generation by 2050 is expected because of the economy modernization.27 Energy systems will show electricity share doubling to exceed 40% of the mix in just two decades. This may affect electricity demand, driving the need for efficient use of distributed energy sources, growing thanks to increased competitiveness of renewables. Cumulative effect of data centers on carbon footprint reduction in power generation and distribution in Figure 5 shows optimistic and pessimistic scenarios leading to reduced emissions in both cases. Stronger sector- specific regulatory hurdles and limited implementation of clean power characterizing pessimistic situation differs from the optimal case. However, in both cases deployment of data centers results with less pollution.

Figure 5 Data centers share in 2050 emissions from power generation sector.

Scope 2 emissions locate power sector as a root cause of the interdependent global carbon footprint. New loads coming from electrification of heating or electric transport, including EV charging in buildings, or other upcoming transformations are going

Chapter 1 – Energy Sector

25 Schneider Electric, Back to 2050 (2021) 26 Energies, A Comprehensive Review of Artificial Intelligence (AI) Companies in the Power Sector (2023) 27 Energy Information Administration (EIA) (2023)

https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp https://www.mdpi.com/1996-1073/16/3/1077 https://www.eia.gov/pressroom/releases/press542.php#:~:text=Compared%20with%202022%2C%20global%20demand,2050%20in%20the%20Reference%20case.

12

to alter demand and induce on-premises power generation and storage. Digital technology will play a key role in building new grid infrastructure and in sectorial integration, mostly through smart management of loads and administration of grid operations on the supply side. Most problems of clean and sustainable power generation will be easier to overcome with smart analytics and machine learning adjusting operations and reporting to frequently changing regulations. Smart algorithms employed in providing demand elasticity will also help to reduce costs and to detect fraudulent activities like meter tampering or billing errors. Cybersecurity in data centers is another bespoke benefit of strengthening IT infrastructure.

Decoupling world energy from the risk of stagnation is the key element of electrification and digitalization of the economy. Data centers are a digital backbone for the upcoming changes bringing a new energy supply paradigm. Already leading in sustainability, data centers will further develop through server optimization and infrastructure adjustments responding to different workloads dedicated to AI training and inference.

Read White Paper 180: "Data Centers are Essential to AI-driven Decarbonization of Power Sector".

https://www.se.com/ww/en/download/document/SPD_WP180_EN/?ssr=true https://www.se.com/ww/en/download/document/SPD_WP180_EN/?ssr=true

13

Manufacturing and production have a substantial share in global emissions, reporting almost one gigaton CO2 annually.28 The impact of industry on the environment can be mitigated by electrification and consecutive digitalization of manufacturing and production. Demand-side measures and energy-efficiency improvements are key elements of decarbonization. Implementation of data center-housed Artificial Intelligence (AI) allows much more precise control in this and other areas. Reduction of emissions will be caused by shifting to renewable power sources, especially in technologies requiring heat, but also from optimizing operations. Industry 4.0 that revolutionizes production through automation and efficiency gains identifies data centers as enablers. Smart digital tools solve the common struggle with IT/OT convergence observed in many sites. Centralized data centers for AI training or data storage are going to coexist with edge infrastructure, mostly used for inference and low-latency low-jitter demanding operations close to data source. Access to local data centers allows Internet of Things (IoT) applications, provides visibility to assets and access to Industrial Automation Control Systems (IACS).29 Solving cybersecurity issues with Industrial Demilitarized Zones (IDMZ) type systems provides safe boundaries between IT and OT domains that have different access levels.

Figure 6 Scenarios for industry emissions reduction with and without data centers.

Industry decarbonization will require investment in industrial sites and production infrastructure including distributed power generation. However, data centers are essential to complete decarbonization of the sector by enabling substantial digitalization levels. Figure 6 shows optimistic and pessimistic scenarios for AI-driven decarbonization of manufacturing. In the coming three decades, data centers can participate in removing 47% more carbon from the environment than without extensive digitalization of production sector. Manufacturing of IT hardware and the operations of data centers have a digital carbon footprint,30 but this impact on the environment is undoubtedly smaller than the emissions saved from the sustainable efficiency that IT brings to industry.

Chapter 2 – Manufacturing Sector

28 Schneider Electric, Back to 2050 (2021) 29 CISCO, Industrial Automation Networks (2021) 30 8 Billion Trees, Carbon Footprint of Data Centers & Data Storage Per Country (2023)

https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp https://www.cisco.com/c/en/us/td/docs/solutions/Verticals/Solution_Briefs/IA_Networking_Solution_Brief.html?DTID=oemels001119&ECID=33344&CCID=cc002683 https://8billiontrees.com/carbon-offsets-credits/carbon-ecological-footprint-calculators/carbon-footprint-of-data-centers/#:~:text=of%20Data%20Centers-,What%20Is%20the%20Carbon%20Footprint%20of%20Data%20Centers%3F,2%25%20of%20global%20GHG%20emissions

14

Socio-economic studies predict that population growth will lead to a requirement for higher efficiency in manufacturing. Personalization of mass production and adaptation of new AI-productivity tools will have a determining impact on IT infrastructure. Massive amounts of data generated in this evolution will be managed in data centers, which will need to be both more sustainable and deployed and in much bigger numbers than today. Data centers will undergo an evolution and adapt to the new distribution of workloads driven by algorithm training and inference. This is described in greater detail in Schneider Electric White Paper 110 “The AI Disruption: Challenges and Guidance for Data Center Design“.

By changing traditional manufacturing process operations and product design, AI will be responsible for the majority of digital information handling and transfer, and it is expected to reach $B253 in a decade only in manufacturing sector.31 New value creation areas like workforce augmentation and broader process automation may not happen without high performance cloud and on premises infrastructure. Besides, productivity growth must involve the full value chain thus facilitating data sharing between ecosystems based on data center cybersecurity and access to IT infrastructure providing boundaries between domains.

Read White Paper 167: "Data centers are Essential to AI-driven Decarbonization of Production and Manufacturing".

31 Precedence Research, Artificial Intelligence Market Size (2023)

https://www.se.com/ww/en/download/document/SPD_WP110_EN/ https://www.se.com/ww/en/download/document/SPD_WP110_EN/ https://www.se.com/ww/en/download/document/SPD_WP167_EN/ https://www.se.com/ww/en/download/document/SPD_WP167_EN/ https://www.precedenceresearch.com/artificial-intelligence-market

15

Transportation interconnects many sectors of the economy, affecting retail supply chains, the manufacturing sector, and safety in cities. Pollution from this widespread sector is significant and forms around 20% of global emissions.32 Evolutionary trends leading to transport decarbonization are thus important and overarching for the economy. Electrification and consecutive digitalization are the two pillars of reducing carbon footprint in transportation. A shift from fossil-based energy sources to much more efficient electricity from renewables requires broad implementation of AI to retain resilient and reliable services. Data analytics like predictive algorithms will help in finding areas for improvement and suggest the most optimal solutions, but this transformation will generate large amounts of data. Sectorial information processed in data centers must address aspects like low latency transfer, cybersecurity, and the availability of compute power. An easy solution to that challenge is broader implementation of centralized and edge data centers.

Expected trends in emissions reduction shown in Figure 7 follow optimistic and pessimistic scenarios.33 In both cases, the complex and multilayered decarbonization challenge requires collective efforts from governments, industry, and individuals. AI has the potential to revolutionize transport by making it more efficient and sustainable. The absence of data centers on the route to clean transport shows a one third higher emissions in either situation, optimistic or pessimistic. Global population increase and the related economy growth unassisted by the IT sector may result in pollution levels higher than observed today.

Figure 7 Emissions in transport sector in different scenarios.

Successful integration of renewables will effectively reduce part of the pollution in transportation. An expected demand shift coming from new generational trends and governmental mitigations of climate change may also play a significant role in decarbonization. Therefore, new models of operations, integration of distributed energy resources, and management of increasing demand pose a challenge for the IT sector.

Chapter 3 – Transport Sector

32 The Eco Experts, Top 4 Most Polluting Industries (2023) 33 Schneider Electric, Back to 2050 (2021)

https://www.theecoexperts.co.uk/blog/top-7-most-polluting-industries https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp

16

Passenger and freight road transport forms nearly three quarters of sectorial emissions.32 These modes are likely to decrease by 2050, while rail, marine, and air are expected to increase. The most pronounced change of 150% demand growth observed for rail suggests that passenger and freight traffic will have to be managed by AI on many levels including energy management, operations efficiency, or assets maintenance. Strategically deployed data centers providing a reliable network that supports decarbonization of transport are crucial in this modern economy.

Read White Paper 168: "Data Centers are Essential to AI-driven Decarbonization of Transport".

https://www.se.com/ww/en/download/document/SPD_WP168_EN/?ssr=true https://www.se.com/ww/en/download/document/SPD_WP168_EN/?ssr=true

17

Addressing pollution from the buildings sector plays a significant role in reducing carbon emissions from the economy.34 The majority of sectorial pollution forecasted in the coming years will come from operations, while the remaining carbon footprint will be associated with embedded carbon. Crucial to reducing the sector’s emissions are AI tools that support building management systems, respect occupancy trends, and control on-site microgrids with smart adjustment of EV charging stations. Digital tools supporting building design and materials management will participate in embedded carbon mitigation in the buildings sector. Fighting climate change will occur with the tremendous growth of data traffic, which will require the heavy presence of local edge and centralized data centers. Buildings will soon become energy centers, creating an integral part of the infrastructure of smart cities. Data centers helping to connect buildings to transportation and other industries like power generation, entertainment, and manufacturing will play a pivotal role in green transformation.

Reduction of sectorial emissions shown in Figure 8 follow two hypothetical scenarios with the optimistic view reporting a 93% emissions reduction by 2050 when data centers will be available. The pessimistic scenario shows an analogous trend with 48% of global emissions removed by 2050 with data centers. Removing IT infrastructure support from the picture sets buildings sector decarbonization behind preset goals leading to a net-zero economy.

Figure 8 Scenarios for reduction of buildings’ emissions observed with and without data centers.

Digitalization of the buildings sector will enable the active monitoring of emissions with identification of pollution reduction opportunities by predictive analytics. Involvement of AI in measuring and reporting pollution will help governments mitigating global carbon footprint and facilitate obligatory reporting for building managers. Connecting IoT sensors into an overarching system of multiple data sources will enable feeding information to smart algorithms. Predictive maintenance tools that indicate potential asset failure will eventually reduce downtime and improve the efficiency and resiliency of buildings.35 Smart systems deployed in decarbonization and efficiency improvement in the buildings sector require more accessibility to data centers. Especially the edge and cloud will allow on-site computation and rapid data analyses leading to process adjustments and better decisions.

Chapter 4 – Buildings Sector

34 Schneider Electric, Back to 2050 (2021) 35 Sensors, Predictive Maintenance in Building Facilities: A Machine Learning-Based Approach (2021)

https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp

18

Read White Paper 169: "Data Centers are Essential to AI-driven Decarbonization of Buildings".

The AI Disruption: Challenges and Guidance for Data Center Design White Paper 110

Back to 2050 1.5ºC is more feasible than we think

REPowerEU: Empowering Energy Consumers for a More Sustainable and Resilient Europe A 10-point action plan to make Europe digital and electric by 2027

The Decarbonization Challenge, Part 1 Closing the Ambition to Action Gap

The Decarbonization Challenge, Part 2 Getting it Done

Browse All White Papers whitepapers.apc.com

Browse All TradeOff Tools™ tools.apc.com

Contact us

For feedback and comments about the content of this report:

Schneider Electric Energy Management Research Center

dcsc@schneider-electric.com

If you are a customer and have questions specific to your data center project:

Contact your Schneider Electric representative at

www.apc.com/support/contact/index.cfm

Resources

https://www.se.com/ww/en/download/document/SPD_WP169_EN/ https://www.se.com/ww/en/download/document/SPD_WP169_EN/ http://https://www.se.com/ww/en/download/document/SPD_WP110_EN/ http://https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/back-to-2050.jsp http:// http:// http://https://perspectives.se.com/e-books/the-decarbonization-challenge-part-1-closing-the-ambition-to-action-gap http://https://perspectives.se.com/e-books/decarbonization-a-holistic-approach-to-organizational-climate-action http://https://www.apc.com/us/en/country-selector/?ref_url=/whitepapers http://https://www.apc.com/us/en/country-selector/?ref_url=/trade-off-tools mailto:dcsc%40schneider-electric.com?subject= http://www.apc.com/support/contact/index.cfm

Executive summary Chapter 1 – The climate imperative calls for a new approach to the energy transition Chapter 2 – In 2050, we will live in a different world