GreenOps Glossary
Cloud and AI sustainability terms, explained
The most talked-about topics in measuring and cutting cloud and AI emissions, explained by the team behind the Greenpixie methodology.
A
AI acceleratorAn AI accelerator is a processor designed to speed up machine learning. GPUs are the most common. Google TPUs and AWS Trainium are purpose-built AI chips.AI carbon footprintAn AI carbon footprint is the greenhouse gas emitted by running and training AI models, from the electricity used and from making the hardware.AWS Cost and Usage Report (CUR)The AWS Cost and Usage Report (CUR) is AWS's most detailed billing export, listing cost and usage line by line for every product and usage type.
C
Carbon accountingCarbon accounting is measuring and reporting the greenhouse gas emissions an organisation causes, usually as a CO2e inventory split into three scopes.Carbon intensityCarbon intensity is the greenhouse gas emitted per unit of electricity consumed from a grid, usually in grams or tonnes CO2e per kWh.CO2eCO2e, or carbon dioxide equivalent, expresses the global warming potential of different greenhouse gases as an equivalent amount of CO2.CSRDThe CSRD is the EU law requiring companies above a certain size to report on sustainability risks and on their impact on people and the environment.
D
Data centre energy consumptionData centre energy consumption is the electricity a data centre draws for its IT equipment plus overhead such as cooling and power distribution.Data centre water usageData centre water usage is the water consumed cooling IT equipment on site, plus the water consumed generating the electricity the site uses.
E
Embodied carbonEmbodied carbon is the greenhouse gas emitted making, transporting and disposing of hardware, separate from the emissions of running it.Emission factorAn emission factor is the amount of greenhouse gas released per unit of activity, such as CO2e per kWh of electricity from a given fuel.
F
G
GHG ProtocolThe GHG Protocol is a family of standards and guidance for measuring and managing greenhouse gas emissions, from WRI and WBCSD.GPUA GPU (graphics processing unit) is a processor with hundreds of cores that run thousands of tasks at once. Most AI training and inference runs on GPUs.GreenOpsGreenOps is the practice of putting energy, carbon and water data next to cost and performance in everyday cloud and AI decisions.
I
L
Large language model (LLM)A large language model (LLM) is a deep learning model trained on very large text datasets to understand and generate language, code and other content.Life cycle assessment (LCA)Life cycle assessment (LCA) measures a product's environmental impacts across every stage of its life, from raw materials to end of life.Location-based vs market-based emissionsLocation-based and market-based are the two GHG Protocol methods for Scope 2: grid-average emissions versus emissions from contracted energy.
O
P
Q
R
Renewable energy certificates (RECs)A renewable energy certificate (REC) is a market-based instrument representing the environmental attributes of 1 MWh of renewable electricity.RightsizingRightsizing is matching a cloud resource's size and type to what its workload actually uses, so you stop paying for idle capacity.
S
Science Based Targets (SBTi)Science-based targets are corporate emissions reduction targets in line with climate science. The SBTi sets the standards and validates them.Scope 1 emissionsScope 1 emissions are direct greenhouse gas emissions from sources an organisation owns or controls, as defined by the GHG Protocol.Scope 2 emissionsScope 2 emissions are indirect emissions from generating the electricity, steam, heat or cooling an organisation buys and uses.Scope 3 emissionsScope 3 emissions are all indirect emissions in an organisation's value chain, upstream and downstream, that are not counted in Scope 2.Sustainable ITSustainable IT is the practice of reducing the environmental impact of all the technology an organisation uses, from laptops to data centres.
T
TokenomicsTokenomics is the economics of AI tokens: what each token costs, how it trades off against performance, and the energy, carbon and water behind it.TokensTokens are the small units of text, such as words or parts of words, that an AI model reads and writes. AI usage and pricing are usually counted in tokens.