diff --git a/_data/carbonStandard.yml b/_data/carbonStandard.yml index e0833bd..6f6ba4c 100644 --- a/_data/carbonStandard.yml +++ b/_data/carbonStandard.yml @@ -22,7 +22,7 @@ title: Technology Carbon Standard CatU: - description: Upstream emissions relating to the embodied carbon of hardware, the development of software, and content. + description: Upstream environmental impacts relating to the embodied carbon and water of hardware, the development of software, and content. CatUSoftware: id: CatUSoftware @@ -85,7 +85,7 @@ CatUContentAndData: noLink: false OperationalEmissions: - description: Operational emissions are associated to the running of business IT functions and systems. Each operational function may have a hybrid of IT infrastructure across direct and/or indirect emissions. + description: Operational environmental impacts are emissions and embodied water associated to the running of business IT functions and systems. Each operational function may have a hybrid of IT infrastructure across direct and/or indirect emissions. CatO: description: Direct running costs that are attributed to the electricity powering servers, networks and devices. @@ -141,7 +141,7 @@ CatGGenerators: noLink: false CatC: - description: Indirect running carbon costs that are attributed to external hardware and service solutions. + description: Indirect running costs that are attributed to external hardware and service solutions. CatCCloud: id: CatCCloud @@ -184,7 +184,7 @@ CatCOffsiteEmployeeDevices: noLink: false CatD: - description: Carbon associated with the use of products and services produced by the business. + description: Carbon and water associated with the use of products and services produced by the business. CatDCustomerDevices: id: CatDCustomerDevices diff --git a/_glossary/blue_water.md b/_glossary/blue_water.md new file mode 100644 index 0000000..a9d885d --- /dev/null +++ b/_glossary/blue_water.md @@ -0,0 +1,7 @@ +--- +title: Blue Water +--- + +Blue Water (BW) refers to freshwater originating from rivers, lakes reservoirs and aquifers. It is used for many purposes, including agricultural, industrial, and domestic. + +It is worth emphasising that most of AI's water footprint comes from blue water, which is directly accessible for human use but often more limited in availability. \ No newline at end of file diff --git a/_glossary/green_water.md b/_glossary/green_water.md new file mode 100644 index 0000000..169c733 --- /dev/null +++ b/_glossary/green_water.md @@ -0,0 +1,5 @@ +--- +title: Green Water +--- + +Green Water (GW) refers to water accessible to plants and microorganisms. It comes from precipitation, which infiltrates and is stored in the soil and can only be used in situ by plants. The main process through which green water moves is "evapotranspiration", where it evaporates from soil into the atmosphere. \ No newline at end of file diff --git a/_glossary/grey_water.md b/_glossary/grey_water.md new file mode 100644 index 0000000..8cf6774 --- /dev/null +++ b/_glossary/grey_water.md @@ -0,0 +1,5 @@ +--- +title: Grey Water +--- + +Grey Water refers to domestic wastewater that comes from sinks, washing machines, bathtubs and showers, exluding toilet wastewater (blackwater). \ No newline at end of file diff --git a/_glossary/water_consumption.md b/_glossary/water_consumption.md new file mode 100644 index 0000000..f93d1a3 --- /dev/null +++ b/_glossary/water_consumption.md @@ -0,0 +1,11 @@ +--- +title: Water Consumption +--- + +Water withdrawal, or water intake, refers to the total volume water withdrawn from its original source. + +Water consumption by contrast is the portion of withdrawn water that is not being returned to its original source, as a result of evaporation, evapotranspiration, product or crop integration, water transfers to different watersheds, consumption or otherwise removed from freshwater resources. + +The difference between water withdrawal and water discharge is the amount consumed. + +By default, water footprint refers to the water consumption unless otherwise specified. \ No newline at end of file diff --git a/_glossary/water_usage_effectiveness.md b/_glossary/water_usage_effectiveness.md new file mode 100644 index 0000000..0654659 --- /dev/null +++ b/_glossary/water_usage_effectiveness.md @@ -0,0 +1,14 @@ +--- +title: Water Usage Effectiveness (WUE) +--- + +A metric used to measure how efficiently data centres use water for cooling and operations. WUE is quantified in cubic meters per megawatt hour of energy (m3/MWh), representing the amount of water consumed per unit of IT equipment output or computing work. +To better understand the true water cost of data centres, source (offsite) and site-based (onsite) WUE metrics must be accounted for. +[The Green Grid](https://airatwork.com/wp-content/uploads/The-Green-Grid-White-Paper-35-WUE-Usage-Guidelines.pdf) distinguishes them as: + +- **WUE**: a site-based metric that is an assessment of the water used on-site for operation of the data +center. This includes water used for humidification and water evaporated on-site for energy production +or cooling of the data center and its support systems (similar to carbon Scope 1). +- **WUEsource**: a source-based metric that includes water used on-site and water used off-site in the +production of the energy used on-site. Typically this adds the water used at the power-generation +source to the water used on-site (similar to carbon Scope 2). diff --git a/index.html b/index.html index 191b590..39d8376 100644 --- a/index.html +++ b/index.html @@ -6,18 +6,25 @@

Technology Carbon Standard

-

The Technology Carbon Standard is a method, structure and data format for analysing an organisation's Technology Carbon Footprint.

+

The Technology Carbon Standard is a method, structure and data format for analysing an organisation's Technology Carbon and Water Footprint.

-

This is developed as an approach to classifying an organisation's technology estate to enable consistent analysis and benchmarking of its technology carbon footprint. This standard takes inspiration from the globally recognised GHG Protocol and its emissions classification into Scopes 1, 2, and 3.

+

This is developed as an approach to classifying an organisation's technology estate to enable consistent analysis and benchmarking of its technology carbon and water footprint. This standard takes inspiration from the globally recognised GHG Protocol and its emissions classification into Scopes 1, 2, and 3.

This method is derived from the Life Cycle Assessment (LCA) methodology, a systematic approach used to evaluate the environmental impact of a product or service through every phase of its life. LCA examines every stage, from the extraction and processing of raw materials through manufacturing, distribution and use, to the disposal or recycling of the product. It also accounts for all associated energy and material flows.

-

In this regard, the Tech Carbon Standard is designed to help organisations link their carbon emissions to specific components of the LCA, enabling more focused analysis and action.

+ +

In this regard, the Tech Carbon Standard is designed to help organisations link their carbon emissions and water usage to specific components of the LCA, enabling more focused analysis and action.

+ +

Water Footprint

+ +

Whilst the tech industry carbon emissions face growing public scrutiny, its water footprint remains largely underlooked despite freshwater scarcity being one of the most pressing environmental and social challenges. The Technology Carbon Standard therefore seeks to bridge this gap by incorporating water as a key environmental impact.

+ +

Assessing both carbon emissions and water consumption present unique challenges as minimising one footprint might increase the other footprint. It is therefore important to adopt a holistic approach that considers both impacts in tandem.

Why a New Standard?

-

The goal of this Technology Carbon Standard is to provide a standardised model of tech emissions that can help organisations when mapping out, measuring or improving the environmental impacts of their technology estate. This fills a gap between high-level consultancy guidance and specific projects tackling individual areas like cloud computing. Consistent standards facilitate productive conversations between sustainability stakeholders, technology leadership and practitioners.

+

The goal of this Technology Carbon Standard is to provide a standardised model of tech emissions and water consumption that can help organisations when mapping out, measuring or improving the environmental impacts of their technology estate. This fills a gap between high-level consultancy guidance and specific projects tackling individual areas like cloud computing. Consistent standards facilitate productive conversations between sustainability stakeholders, technology leadership and practitioners.

-

By delineating an organisation's technology estate into categories analogous to GHG Protocol scopes, technology practitioners can more easily identify the most carbon-intensive areas to prioritise for impact reduction. This is not intended to reinvent existing efforts in sustainability measurement and reporting. Where relevant, the methodology references and integrates with established open-source initiatives for calculating technology emissions.

+

By delineating an organisation's technology estate into categories analogous to GHG Protocol scopes, technology practitioners can more easily identify the most carbon and water intensive areas to prioritise for impact reduction. This is not intended to reinvent existing efforts in sustainability measurement and reporting. Where relevant, the methodology references and integrates with established open-source initiatives for calculating technology emissions and water consumption.

@@ -26,7 +33,7 @@

Why a New Standard?

-

Note that the Tech Carbon Standard does not include other aspects of an organisation's carbon footprint. For example, People, Buildings and Travel (i.e. commuting) are not included in the standard.

+

Note that the Tech Carbon Standard does not include other aspects of an organisation's carbon and water footprints. For example, People, Buildings and Travel (i.e. commuting) are not included in the standard.

This standard provides comprehensive coverage of an organisation's technology carbon footprint while clarifying the different sources of emissions. Improving quantification and transparency aims to spur more targeted, impactful efforts to reduce tech-related emissions.

diff --git a/pages/about.md b/pages/about.md index b98d8bf..9729a7c 100644 --- a/pages/about.md +++ b/pages/about.md @@ -6,11 +6,11 @@ permalink: /about # About -The Technology Carbon Standard provides a structure for organisations to understand, measure, and reduce the carbon emissions associated with their technology infrastructure and usage. It is part of a broader Sustainable Technology Framework that incorporates additional sustainability considerations beyond just carbon. +The Technology Carbon Standard provides a structure for organisations to understand, measure, and reduce the carbon emissions and water consumption associated with their technology infrastructure and usage. It is part of a broader Sustainable Technology Framework that incorporates additional sustainability considerations beyond just carbon. -Specifically, the Carbon Standard focuses on quantifying and minimising the carbon footprint of an organisation’s technology estate. +Specifically, the Carbon Standard focuses on quantifying and minimising the carbon and footprints of an organisation’s technology estate. -By providing a structure to categorise and account for carbon across these different technology facets, the Standard aims to help organisations map out their technology footprint, identify hotspots for reduction opportunities, and track progress over time. It serves as a proposed standard for organisations on their journey toward sustainable technology. Adopting the Technology Carbon Standard brings organisations one step closer to fully understanding and minimising the climate impacts of their digital ecosystem. +By providing a structure to categorise and account for carbon and water across these different technology facets, the Standard aims to help organisations map out their technology footprint, identify hotspots for reduction opportunities, and track progress over time. It serves as a proposed standard for organisations on their journey toward sustainable technology. Adopting the Technology Carbon Standard brings organisations one step closer to fully understanding and minimising the climate impacts of their digital ecosystem. The Technology Carbon Standard was developed and is maintained by [Scott Logic](https://www.scottlogic.com) and offered as an open-source project. Scott Logic has aligned its approach with other open-source and science-backed, cross-industry frameworks such as the GHG Protocol and Green Software Foundation. diff --git a/pages/impact_categories/downstream.md b/pages/impact_categories/downstream.md index 895fbbe..56c73d0 100644 --- a/pages/impact_categories/downstream.md +++ b/pages/impact_categories/downstream.md @@ -10,6 +10,8 @@ redirect_from: Downstream emissions (Category D) are the emissions produced by using an organisation’s products and/or services. This could be B2B or B2C users of the products and services. The emissions are attributed to the customer’s device energy use and the transmission of data to use that product or service. +To assess the downstream water footprint of an organisation's operations, it is essential to include the water required to use the products or services, as well as the water embedded in the energy consumed by customer devices and data transmission. + Downstream emissions are related to GHG Protocol Scope 3. {% include linkedHeading.html heading="Customer Devices" level=2 %} @@ -18,15 +20,17 @@ This considers the emissions generated from the electricity consumption of devic The [embodied carbon of customer devices](/technology-categories/lifecycle/embodied) is not required for attributional accounting of indirect scope 3 emissions. However, there is a consequential effect that organisations should consider when releasing digital products and services. If the software requires significantly powerful hardware or requires the latest operating systems, for example, this could force customers to replace their hardware and devices to be able to use the product. This carries a downstream consequence of increasing the embodied emissions of customer devices, as well as increasing waste. [Read more about hardware lifecycle emissions](/technology-categories/lifecycle). -When content and data are consumed by customers, it is essential to account for emissions generated across all interactions, including: +The embodied or virtual water tied to customer devices includes both the water consumed to generate the electricity they consume and the water embedded in their manufacturing. As software requires more computing power, it increases electricity consumption. In addition, when new software requires customers to upgrade or replace their devices, the water impact grows further. + +When content and data are consumed by customers, it is essential to account for both the carbon and water usages across all interactions, including: - Streaming, playback or download - Content creation such as live streaming and uploads - Embedding and sharing on third-party platforms - AI-enabled features such as auto-captions and recommendations -Special consideration also needs to be given to AI applications. As the size of LLMs continues to increase so do the hardware requirements needed to run them effectively. If your customers need to upgrade their devices to run your AI applications this will contribute to e-waste as mentioned above and could also lead to increased energy consumption. +Special consideration also needs to be given to AI applications. As the size of LLMs continues to increase so do the hardware requirements needed to run them effectively. If your customers need to upgrade their devices to run your AI applications this will contribute to e-waste as mentioned above and could also lead to increased energy and water consumption. -It is also important to consider the number of end users that will be consuming your AI. A [study released by Mistral AI in July 2025](https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai) into the energy usage of their Large 2 model stated that to generate 1 page of text or 400 tokens produces 1.14 gCO₂e. This is roughly equivalent to a user in the U.S. streaming online video for 10 seconds. While this per-request impact may appear minimal, the cumulative emissions scale rapidly when multiplied across thousands of users performing inferences multiple times per day. +It is also important to consider the number of end users that will be consuming your AI. A [study released by Mistral AI in July 2025](https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai) into the energy usage of their Large 2 model stated that to generate 1 page of text or 400 tokens produces 1.14 gCO₂e and uses 45 mL of water. This is roughly equivalent to a user in the U.S. streaming online video for 10 seconds. While this per-request impact may appear minimal, the cumulative emissions and water consumed scale rapidly when multiplied across thousands of users performing inferences multiple times per day. It is therefore important to consider ways to reduce the number of inference attempts a user needs to obtain the desired information as well as reducing how long inference takes and the length of the answer returned. @@ -46,8 +50,8 @@ In the context of AI applications, writing more precise prompts and requesting c These emissions come from physical infrastructure and systems that your customers deploy or operate to use your products or services. This includes IoT (Internet of Things) devices, locally hosted servers, edge computing / embedded devices, dedicated storage systems, networking equipment, and specialised hardware (also know as OT or Operational Technology) that customers must install for your service to function. For example, if your cloud-based IoT platform requires customers to install local gateways and sensors, or your enterprise software needs on-premises servers at customer sites. -The usage related emissions from this customer infrastructure should be accounted for under the customer's operational energy consumption (unless it's something like a managed service where you pay for the electricity consumed - in which case it should be included under your operational emissions) and, where relevant, the embodied carbon of hardware that customers acquire specifically to use your offerings. Unlike general-purpose devices, this infrastructure often has longer lifecycles and a range of power requirements (depending on the nature of the device). +The usage related emissions from this customer infrastructure should be accounted for under the customer's operational energy and therefore water consumption which can vary greatly depending on the type energy used (unless it's something like a managed service where you pay for the electricity consumed - in which case it should be included under your operational emissions) and, where relevant, the embodied carbon of hardware that customers acquire specifically to use your offerings. Unlike general-purpose devices, this infrastructure often has longer lifecycles and a range of power requirements (depending on the nature of the device). -Consider the cumulative impact across your entire customer base, especially where your service requires distributed infrastructure deployment. The geographic spread of customer infrastructure affects carbon intensity calculations, as different regions have varying grid emissions. Your product design decisions around hardware requirements, efficiency, and deployment patterns directly influence the overall carbon footprint of your service ecosystem. +Consider the cumulative impact across your entire customer base, especially where your service requires distributed infrastructure deployment. The geographic spread of customer infrastructure affects carbon intensity and water consumption calculations, as different regions have varying grid emissions. Your product design decisions around hardware requirements, efficiency, and deployment patterns directly influence the overall carbon and water footprints of your service ecosystem. If your AI application requires your customers to host the application locally then they may need to implement specialised hardware to ensure optimal performance. For example, dedicated servers and GPUs, additional RAM, specialist storage solutions such as NVMe SSDs, and enhanced cooling systems may all be necessary. \ No newline at end of file diff --git a/pages/impact_categories/operational.md b/pages/impact_categories/operational.md index ef30cfa..c641752 100644 --- a/pages/impact_categories/operational.md +++ b/pages/impact_categories/operational.md @@ -7,9 +7,9 @@ redirect_from: --- # Operational Emissions -Operational emissions encompass the greenhouse gases emitted from an organisation's own technology infrastructure and operations. These are categorised into three groups - On-Prem, Cloud, and Generator emissions (O, C, and G) - based on the level of ownership and control an organisation has over the assets. +Operational emissions encompass the greenhouse gases emitted from an organisation's own technology infrastructure and operations as well as the water consumption involved in the latter. The grenhouse gases emissions are categorised into three groups - On-Prem, Cloud, and Generator emissions (O, C, and G) - based on the level of ownership and control an organisation has over the assets. -Understanding operational emissions allows organisations to quantify, monitor, and benchmark the climate impacts of their technology infrastructure and business operations. It enables setting emission reduction targets and strategies and facilitates compliance with current and emerging climate disclosure legislations. +Understanding operational emissions and water consumption allows organisations to quantify, monitor, and benchmark the climate impacts of their technology infrastructure and business operations. It enables setting emission reduction targets and strategies and facilitates compliance with current and emerging climate disclosure legislations. Emissions are further classified into direct and indirect sources: @@ -17,6 +17,8 @@ Emissions are further classified into direct and indirect sources: Direct emissions are a result of the organisation's direct consumption of grid-supplied electricity (Category O) or combustion of fossil fuels like diesel or natural gas for owned power generators (Category G). The organisation can directly measure and account for emissions from owned assets. +Moreover, grid-supplied electricity consumption is closely linked to water usage, as different energy sources vary significantly in their water requirements. + {% include linkedHeading.html heading="On-premise" level=3 %} Emissions associated with the actual operation and use of devices owned by an organisation ([usage carbon](/resources/glossary#usage-carbon)). @@ -29,6 +31,8 @@ Category O (On-premise) emissions can be related to GHG Protocol Scope 2. The energy consumed by on-premise servers and data centres. This includes those dedicated to AI operations and associated cooling systems, and other infrastructure necessary for maintaining AI workloads. +Servers and data centres consume vast amounts of water, primarily to cool their processor chips and to avoid overheating (onsite water). But the true water cost of data centres also includes the water consumed to generate the electricity that powers them (offsite water). To fully evaluate their water impact, both sources must be considered. + {% include linkedHeading.html heading="Machine Learning training and fine-tuning" level=5 %} {% include linkedHeading.html heading="Foundation Models" level=6 %} @@ -37,7 +41,7 @@ This section focuses on organisations that develop and train foundation models, These operations typically demand intensive use of specialised hardware including Graphics Processing Units (GPUs) and Central Processing Units (CPUs), along with associated infrastructure such as high-performance storage systems, large amounts of RAM and cooling equipment. -Assessing the efficiency of hardware and algorithms can help reduce unnecessary energy consumption. Techniques such as model distillation, quantisation, energy-aware pruning and low-precision arithmetic operations can further lower the computational and energy requirements of models. +Assessing the efficiency of hardware and algorithms can help reduce unnecessary energy and water consumption. Techniques such as model distillation, quantisation, energy-aware pruning and low-precision arithmetic operations can further lower the computational and energy requirements of models. It should be acknowledged however, that not all model development requires vast amounts of computing power: while Large Language Models are resource-intensive and versatile, smaller models (SLMs) trained on more modest datasets are designed to be more compact and efficient, requiring less computational power and memory. @@ -85,6 +89,8 @@ Special consideration also needs to be given to the increased use of LLMs. As th Any fossil fuel-powered generators, solar PV, wind turbines, or other systems installed on-site to supply electricity to technology equipment. +Any generators that use water to generate electicity such as hydroelectric power generators must also be accounted for. + Category G (Generators) can be related to GHG Protocol Scope 1. {% include linkedHeading.html heading="Indirect Emissions" level=2 %} @@ -98,6 +104,8 @@ Category C (Cloud) emissions can be related to GHG Protocol Scope 3. The emissions associated with cloud platform services like compute, storage and networking. Services are backed by computing hardware with associated upstream and operational emissions. The proportion of such emissions attributable to an organisation will vary based on service, server instance types, and region. +The water consumed to cool down servers but also indirectly to generate the electricity to power centres. + If your organisation is developing an AI product — whether you're building a large language model from the ground up or leveraging an existing foundation model — it's highly likely that it will be hosted on a [Cloud platform](/resources#ai-cloud-providers), whether using Model as a Service (Maas) or self-managed Cloud infrastructure for custom deployments. This is due to the substantial computational resources required to run AI systems efficiently. [Read more about cloud services.](/technology-categories/cloud) diff --git a/pages/impact_categories/upstream.md b/pages/impact_categories/upstream.md index 2b2f9f8..bb9036e 100644 --- a/pages/impact_categories/upstream.md +++ b/pages/impact_categories/upstream.md @@ -10,7 +10,7 @@ redirect_from: Upstream emissions (Category U) refer to the [embodied carbon](/resources/glossary#embodied-carbon) emissions of hardware purchased by an organisation and the carbon emissions of the development and distribution of installed software used by the organisation. -Understanding upstream emissions allows organisations to comprehensively assess the environmental impact of their assets and supply chains. It enables the development of strategies to reduce the carbon footprint of materials and processes, promoting sustainable procurement practices and circular economy principles. +Evaluating both carbon emissions and water consumption to understand upstream emissions allows organisations to comprehensively assess the environmental impact of their assets and supply chains. It enables the development of strategies to reduce the carbon and water footprints of materials and processes, promoting sustainable procurement practices and circular economy principles. Upstream emissions are related to GHG Protocol Scope 3. @@ -24,7 +24,7 @@ Emissions associated with developing and delivering off-the-shelf and open-sourc {% include linkedHeading.html heading="Hardware" level=2 %} -Embodied carbon emissions associated with hardware devices owned by an organisation, including emissions from the manufacture, transportation, installation, maintenance, and end-of-life of a device. +Embodied carbon emissions and water consumption associated with hardware devices owned by an organisation, including emissions from the manufacture, transportation, installation, maintenance, and end-of-life of a device. [Read more about embodied carbon emissions.](/technology-categories/lifecycle/embodied) @@ -32,11 +32,15 @@ Embodied carbon emissions associated with hardware devices owned by an organisat Laptops, desktops, mobiles, printers, and peripherals used by employees. +To accurately assess their emissions, the whole supply chain must be examined, from raw material extraction to manufacturing. For example, semiconductor production relies on energy-intensive techniques, contributing substantially to carbon emissions. + +The production of these electronic devices is also extremely water-intensive. Their water footprint stems from the virtual water embedded in every stage of global manufacturing, from mining precious metals and producing synthetic chemicals for glue and plastic, to assembling and packaging. + [Click here to see a worked example of estimating embodied emissions for a laptop.](/technology-categories/lifecycle/example/employee#embodied-carbon-emissions) {% include linkedHeading.html heading="Networking Hardware" level=3 %} -When considering the upstream emissions of a network, consider the embodied carbon of any networking devices that are owned by the organisation. These include, but are not limited to: +When considering the upstream emissions of a network, consider the embodied carbon and water of any networking devices that are owned by the organisation. These include, but are not limited to: - routers - switches - bridges @@ -55,6 +59,13 @@ Servers, storage systems, and data centre infrastructure installed on-premise. [Click here to see a worked example of estimating embodied emissions for a server.](/technology-categories/lifecycle/example/server#embodied-carbon-emissions) +Assessing the water footprint of data centres requires examining three key areas: +- Onsite water: used to power and cool the IT infrastructure +- Embodied water: involved in server design and manufacture +- Offsite water: required to generate electricity +While renewable sources generate fewer carbon emissions, they may need propertionally larger amounts of water to generate electricity. This trade-off highlights the need to co-optimise carbon and water footprints. +The rapid expansion of AI workloads in data centres intensifies this challenge, due to their growing reliance on water-intensive cooling technologies to meet computing demands. + {% include linkedHeading.html heading="Content" level=2 %} Embodied carbon emissions associated with producing, distributing and storing digital content. @@ -62,7 +73,7 @@ Embodied carbon emissions associated with producing, distributing and storing di {% include linkedHeading.html heading="Foundation Models" level=3 %} AI workloads rely on specialised hardware and AI accelerators to perform complex computations at high speed. -When assessing the upstream emissions of foundation models, it's essential to account for the embodied carbon of the IT infrastructure involved in their training, deployment and with inference. This includes the carbon footprint associated with manufacturing specialised hardware and constructing and operating AI data centres. Key components include, but are not limited to: +When assessing the upstream emissions of foundation models, it's essential to account for the embodied carbon and water of the IT infrastructure involved in their training, deployment and with inference. This includes the carbon footprint and water consumption associated with manufacturing specialised hardware and constructing and operating AI data centres. Key components include, but are not limited to: - High-Performance Computing systems (HPC) - Graphics Processing Units (GPUs) - Central Processing Units (CPUs) diff --git a/schemas/reporting_organisation/v0.0.1.json b/schemas/reporting_organisation/v0.0.1.json index 002bf73..10be671 100644 --- a/schemas/reporting_organisation/v0.0.1.json +++ b/schemas/reporting_organisation/v0.0.1.json @@ -122,7 +122,7 @@ }, "upstream_emissions": { "type": "object", - "description": "Upstream emissions relating to the embodied carbon of hardware, the development of software, and content.", + "description": "Upstream environmental impacts relating to the embodied carbon and water of hardware, the development of software, and content.", "properties": { "software": { "$ref": "#/$defs/emissions_def", @@ -170,7 +170,7 @@ }, "indirect_emissions": { "type": "object", - "description": "Indirect running carbon costs that are attributed to external hardware and service solutions.", + "description": "Indirect running costs that are attributed to external hardware and service solutions.", "properties": { "offsite_employee_hardware": { "$ref": "#/$defs/emissions_def", diff --git a/schemas/tech_carbon_standard/v0.0.1.json b/schemas/tech_carbon_standard/v0.0.1.json index 4d6fec3..f616754 100644 --- a/schemas/tech_carbon_standard/v0.0.1.json +++ b/schemas/tech_carbon_standard/v0.0.1.json @@ -50,7 +50,7 @@ }, "upstream_emissions": { "type": "object", - "description": "Upstream emissions relating to the embodied carbon of hardware, the development of software, and content.", + "description": "Upstream environmental impacts relating to the embodied carbon and water of hardware, the development of software, and content.", "properties": { "software": { "$ref": "#/$defs/emissions_def", @@ -98,7 +98,7 @@ }, "indirect_emissions": { "type": "object", - "description": "Indirect running carbon costs that are attributed to external hardware and service solutions.", + "description": "Indirect running costs that are attributed to external hardware and service solutions.", "properties": { "offsite_employee_hardware": { "$ref": "#/$defs/emissions_def", diff --git a/schemas/tech_carbon_standard/v0.0.2.json b/schemas/tech_carbon_standard/v0.0.2.json index d36dd45..62085f8 100644 --- a/schemas/tech_carbon_standard/v0.0.2.json +++ b/schemas/tech_carbon_standard/v0.0.2.json @@ -50,7 +50,7 @@ }, "upstream_emissions": { "type": "object", - "description": "Upstream emissions relating to the embodied carbon of hardware, the development of software, and content.", + "description": "Upstream environmental impacts relating to the embodied carbon and water of hardware, the development of software, and content.", "properties": { "software": { "$ref": "#/$defs/emissions_def", @@ -98,7 +98,7 @@ }, "indirect_emissions": { "type": "object", - "description": "Indirect running carbon costs that are attributed to external hardware and service solutions.", + "description": "Indirect running costs that are attributed to external hardware and service solutions.", "properties": { "offsite_employee_hardware": { "$ref": "#/$defs/emissions_def",