Introduction Empowering individuals Working together Operating sustainably Governance and engagement Appendix 2023 Environmental Repo爀琀 third-pa爀琀y data centers. The 2022 data center CFE number represents a Google Search interest since 2004. distribution) to the amount of energy used to power IT equipment. A PUE of 2% change from 2021 (66% CFE), as a result of including third-pa爀琀y data 46 See endnote 35 above. 2.0 means that for every wa琀琀 of IT power, an additional wa琀琀 is consumed Endnotes centers. 47 See endnote 43 above. to cool and distribute power to the IT equipment. A PUE closer to 1.0 means 13 nearly all the energy is used for computing. Based on Google Trends data as of December 2022, when comparing 48 See endnote 35 above. 77 global Google Search interest since 2004. 49 “Measuring Fashion: Environmental Impact of the Global Apparel and According to the Uptime Institute’s 2022 Global Data Center Survey, the 14 See endnote 3 above. Footwear Industries Study,” Quantis, 2018. global average PUE of respondents’ data centers was around 1.55. 1 Throughout this repo爀琀, we use the term “sustainability” to refer to 50 78 15 See endnote 6 above. “Digital Transformation,” UNEP, April 2022. See endnote 19 above. environmental sustainability. 16 See endnote 7 above. 51 “Digital Decarbonisation: How the Digital Sector is Suppo爀琀ing Climate 79 Trailing 12-month PUE consists of the average PUE for the previous 12 2 Our pa爀琀ners include governments and intergovernmental organizations; 17 See endnote 8 above. Action,” Implement Consulting Group, 2022. months. customers and commercial pa爀琀ners; and researchers, academics, and 52 18 Based on Google’s analysis of tra昀케c pa琀琀erns before and a昀琀er “68% of the World Population Projected to Live in Urban Areas by 2050, 80 Data from repo爀琀ed po爀琀s for ChargePoint stations in the United States and non-governmental organizations. recommended adjustments to tra昀케c signals that were implemented during Says UN,” United Nations Depa爀琀ment of Economic and Social A昀昀airs, May Canada only are used to estimate emissions prevented from employee EV 3 Google uses a high-quality ML prediction model to estimate the expected 2018. tests conducted in 2022. commuting. fuel or energy consumption for each route option when users request 19 53 81 According to Google’s own analysis of our more e昀케cient servers, power SDG Indicators, United Nations: Unstats, 2022. Based on data collected from shu琀琀le commuting trips in the Bay Area, driving directions. We identify the route that we predict will consume the infrastructure, and cooling systems, compared with data center industry 54 “Izmir Sustainable Energy Climate Action Plan,” AECOM, 2020. as compared to a scenario in which these employees had each used a least amount of fuel or energy. If this route isn’t already the fastest one and averages. 55 See endnote 18 above. private, fuel-based car for commuting. Assumptions include average it o昀昀ers meaningful energy and fuel savings with only a small increase in 20 Googler commute distance, miles per gallon assumptions from the Bureau According to Google’s pla琀昀orm-neutral measurement for central 56 “Spreading Like Wild昀椀re: The Rising Threat of Extraordinary Landscape driving time, we recommend it to the user. To calculate enabled emissions of Transpo爀琀ation, and emissions per gallon of fuel assumptions from the processing unit (CPU) resources analyzed over a 昀椀ve-year period. Fires,” UNEP, 2022. reductions, we tally the fuel usage from the chosen fuel-e昀케cient routes 21 For more information on how we’re reducing the environmental footprint of 57 Our “Deep Learning for High-Resolution Wild昀椀re Modeling” research was Greenhouse Gas Protocol. Equivalency estimate is based on workdays in and subtract it from the predicted fuel consumption that would have the year, using data from “Greenhouse Gas Equivalencies Calculator,” U.S. our operations, see the Operating sustainably section. presented at the 2022 International Conference for Forest Fire Research. occurred on the fastest route without eco-friendly routing and apply 58 Environmental Protection Agency, accessed 2022. 22 “Japan’s Greenhouse Gas Emissions Rose 2% in FY21/22 as Economy “Number of deaths due to 昀氀oods worldwide from 1960 to 2020,” Statista, adjustments for factors such as: CO e factors, 昀氀eet factors, well-to-wheels 82 2 Recovered,” Reuters, April 2023. April 2023. See endnote 81 above. factors, and powe爀琀rain mismatch factors. We then input the estimated 23 59 See endnote 13 above. “Pa爀琀nering with Member States,” UNEP, 2021. 83 Carbon-free energy is any type of electricity generation that doesn’t prevented emissions into the EPA’s Greenhouse Gas Equivalencies 24 “Net Zero by 2050 - A Roadmap for the Global Energy Sector,” IEA, October 60 Pitney Bowes Parcel Shipping Index, 2021. directly emit carbon dioxide, including (but not limited to) solar, wind, Calculator to calculate equivalent cars o昀昀 the road for a year. The 2022 2021. 61 “Retail Speaks,” McKinsey & Company, 2021. geothermal, hydropower, and nuclear. Sustainable biomass and carbon 昀椀gure covers estimated emissions prevented a昀琀er eco-friendly routing was 25 2 capture and storage (CCS) are special cases considered on a case-by-case “CO and Greenhouse Gas Emissions,” OurWorldInData.org, 2020. 62 “Feasibility of A昀昀orestation as an Equitable Nature-Based Solution in Urban launched, from October 2021 through December 2022. Enabled emissions 26 Based on Google Trends data as of December 2022, when comparing Areas,” Sustainable Cities and Society, vol. 81, June 2022. basis, but are o昀琀en also considered carbon-free energy sources. reductions estimates include inherent unce爀琀ainty due to factors that 63 84 global Google Search interest since 2013. Based on Google Trends data as of December 2022, when comparing Alphabet’s renewable energy methodology is a custom calculation and is include the lack of primary data and precise information about real-world 27 See endnote 13 above. global Google Search interest in 2022 to that of 2021. based on a global approach. Percentage of renewable energy is calculated actions and their e昀昀ects. These factors contribute to a range of possible 28 See endnote 6 above. 64 See endnote 35 above. on a calendar-year basis, dividing the volume of renewable electricity outcomes, within which we repo爀琀 a central value. 29 See endnote 7 above. 65 Some of our locations have more than one data center campus, and others (in megawa琀琀-hours) procured for our global operations (i.e., renewable 4 This estimated spend is based on contracts signed to purchase clean 30 See endnote 8 above. weren’t yet operational during 2022. energy procured through our PPA contracts, on-site renewable energy energy for our operations. Actual spend may vary from these estimates 31 See endnote 6 above. 66 “Climate Change 2022: Impacts, Adaptation and Vulnerability generation, and renewable energy in the electric grids where our facilities based on changes in renewable electricity output from operational 32 Nest Renew Energy Shi昀琀 usage re昀氀ects the total number of hours Working Group II Contribution to the Sixth Assessment Repo爀琀 of the are located) by the total volume of electricity consumed by our global projects, the number of contracts signed, and energy market prices. operations. The numerator includes all renewable energy procured, Intergovernmental Panel on Climate Change,” IPCC, 2022. thermostats enrolled in Nest Renew ran Energy Shi昀琀 features to prioritize 5 See endnote 3 above. cleaner or cheaper energy during 2022. 67 “Climate Commitments Not On Track to Meet Paris Agreement Goals regardless of the market in which the renewable energy was consumed. 6 Estimated energy savings are calculated based on the typical percentage 33 Additional details on Alphabet’s criteria and methodology can be found in Nest Renew users include anyone in the continental U.S. with a 3rd as NDC Synthesis Repo爀琀 is Published,” United Nations Climate Change, of heating and cooling savings found in real-world studies of the Nest the “Achieving Our 100% Renewable Energy Purchasing Goal and Going February 2021. generation Nest Learning Thermostat, the Nest Thermostat E, or the Learning Thermostat. To calculate the total Nest savings, we apply these 68 Beyond” disclosure. newest Nest Thermostat connected to a Google account. “Is Carbon Capture Too Expensive?” IEA, February 2021. savings percentages to the actual heating and cooling hours of all Nest 34 85 See endnote 6 above. 69 See the Environmental data tables for Carbon intensity per unit of revenue For example, research by Princeton University and TU Berlin has shown that thermostats. 35 Based on Google Trends data as of December 2022, when comparing for 2022 (9.13 tCO2e/million U.S.$). Our 2011 carbon intensity factor was hourly (24/7) carbon-free energy goals have a greater impact on grid-level 7 “Po爀琀ugal - Countries & Regions,” IEA, 2021. decarbonization than the prevailing approach of 100% annual renewable global Google Search interest since 2008. repo爀琀ed in prior Environmental Repo爀琀s as 38.8 (tCO2e of gross global 8 The enabled emissions reductions are calculated based on these energy 36 See endnote 13 above. combined Scope 1 and market-based Scope 2 emissions/million U.S.$). energy matching. savings, applying standard emission factors for fossil fuels, and using U.S. 70 86 37 See endnote 35 above. This third-pa爀琀y projection is based on a 2020 whole-building LCA, See endnote 4 above. EPA AVERT marginal emissions for the 95% of electricity savings that occur 38 “Transpo爀琀: Improving the Sustainability of Passenger and Freight following the LEED v4 Building Life Cycle Impact Reduction requirements. 87 See endnote 4 above. in the U.S., with an adjusted value for the 5% of electricity savings outside Transpo爀琀,” IEA, 2022. The interior design was not factored into this LCA. 88 “Greenhouse Gas Equivalencies Calculator,” EPA, accessed 2022. the U.S. 39 71 Based on Google Trends data as of April 2023, when comparing global Refrigerants used in HVAC and refrigeration equipment are o昀琀en synthetic 89 Our CFE percentage measures the degree to which our electricity 9 For products launched in 2022: Pixel, Nest, and Chromecast devices are Google Search interest from April 2022–April 2023 to that of April 2021– compounds commonly referred to as super pollutants because their global consumption on a given regional grid is matched with CFE on an hourly made with recycled materials ranging between 9% and 60% based on April 2022. warming potential is signi昀椀cantly higher than other GHGs like carbon basis. This is calculated using both CFE under contract by Google as well respective product weights. 40 dioxide. as CFE coming from the overall grid mix. CFE coming from the overall grid See endnote 3 above. 10 Unique, signed-in Google users that were provided information to make a 72 mix is based on data obtained from a third pa爀琀y, Electricity Maps, and has 41 “Global EV Outlook 2023 Abstract,” IEA, April 2023. “The Carbon Footprint of Machine Learning Training Will Plateau, Then more sustainable choice by at least one sustainable product feature. 42 See endnote 13 above. Shrink,” Computer, vol. 55, July 2022. not been assured. For more information, see our 2021 white paper, “24/7 11 Due to a methodology change, we’re not repo爀琀ing a percent progress 73 Carbon-Free Energy: Methodologies and Metrics.” 43 Based on Google Trends data as of April 2023, when comparing global See endnote 20 above. towards this target and have instead provided our total GHG emissions 90 Google Search interest from April 2020–April 2023 to that of April 2017– 74 See endnote 19 above. A grid region (or regional grid) corresponds to the area over which a (Scope 1, 2 (market-based), and 3) in 2022. April 2020. 75 See endnote 20 above. single entity manages the operation of the electric power system and 12 In 2022, we expanded our carbon-free energy repo爀琀ing from only 44 See endnote 13 above. 76 PUE is a standard industry ratio that compares the amount of non- ensures that demand and supply are 昀椀nely balanced. In the United States, Google-owned and -operated data centers to also incorporate o昀케ces and 45 Based on Google Trends data as of April 2023, when comparing global computing overhead energy (used for things like cooling and power this generally means the ISO or RTO in regions that have these regional 101