This dashboard displays the data sources that Tailpipe uses to generate its carbon emissions calculations. In the interest of transparency, Tailpipe has scored each source on its accuracy, completeness, credibility, conciseness, latency, and plausibility. Tailpipe uses the most up to date and valid data sources available, and has highlighted any improvements to the current data set that it is working to implement in the To Do column.

Key

Accuracy	Valid, replicable, and reliable.	Some gaps in validity, replicability, and reliability, but margins for error have been identified.	Gaps in validity, replicability, and reliability, and margins for error cannot be identified.
Completeness	Fulfills all parameters.	Does not fulfill all parameters, but margins for error have been identified.	Does not fulfill all parameters and margins for error cannot be identified.
Credibility	Source is recognized and has been published according to scientific standards (i.e. a journal or research group) or directly from cloud provider.	Source is recognized but has not been published according to scientific standards (i.e. a brand/non-profit).	Source is not recognized and cannot be traced.
Latency	Methodology sources – Less than 2 years old. Variable sources – less than 1 month.	Methodology sources – 2-5 years old. Variable sources – less than 2 months.	Methodology sources – More than 5 years old. Variable sources – more than 2 months.
Plausibility	Aligns with Tailpipe’s reasonable assumptions.	Sits outside of Tailpipe’s reasonable assumptions.	Exceeds Tailpipe’s assumptions significantly.

Amazon Web Services

Updated: 25/11/24

AWS Operational Emissions Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
CPU Utilisation	Cloudwatch	Accuracy	Completeness	Credibility	Latency	Plausibility		Customer must pay for 5 minute updates when making more than 1 million API requests.
Billing & Usage	Cost and Usage Reports	Accuracy	Completeness	Credibility	Latency	Plausibility		AWS retroactively changes the contents of the CUR in line with new data.
CPU Power	Teads CPU Power Extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	AMI testing to validate that Teads data is valid.	Based on Teads Engineering testing of 6 instances and extrapolating to all CPUs.
GPU Power	Teads GPU Power Draw	Accuracy	Completeness	Credibility	Latency	Plausibility	Needs GPU power sensing on user side – currently no live data.	Limited dataset.
RAM Power	Teads testing, plus 5% increment extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	Lab testing to verify Teads’ RAM figures.
SSD Power	13 drives with capacities from 0.5TB to 8TB (See Appendix B)	Accuracy	Completeness	Credibility	Latency	Plausibility	Assess a wider range of SSD datasheets.
HDD Power	71 drives with capacities from 1TB to 24TB (See Appendix B)	Accuracy	Completeness	Credibility	Latency	Plausibility
Motherboard Power 42.5W	A1 Electronics	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a published study on motherboard consumption.	No demonstrable evidence behind A1 Electronics source, though its statement aligns with other anecdotal evidence.
NICs Power	2017 GeekWire Article and NIC manufacturers	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a more recent source for GBps speed.	Unknown which NICs CSPs use.
Network Storage Power	2017 GeekWire Article and assumption about 16kWh power draw	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a more recent source than GeekWire, and better insight into AWS storage server equipment.	Based on assumptions and 2017 source.
Intra- and Inter-region Networking	Ficher et al., 2021	Accuracy	Completeness	Credibility	Latency	Plausibility	Find an inter-region specific study.	Research paper only analyzed networking within one country.
External Networking	Collected research papers, 2014-2021 (See Appendix E)	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a more recent study to validate these results.
Power Supply Efficiency Factor 1.1	2017 GeekWire Article and PSU manufacturers	Accuracy	Completeness	Credibility	Latency	Plausibility		2017 GeekWire article is outdated but manufacturer data provides credibility.
PUE 1.22	AWS 2024 blog post citing 2023 IDC report	Accuracy	Completeness	Credibility	Latency	Plausibility
Grid Mix – Non-UK	Ember	Accuracy	Completeness	Credibility	Latency	Plausibility	Waiting for Ember to update their dataset to daily or hourly measurements.	1 month latency for every country. Unknown which datacenters generate their own power
Grid Mix – UK	National Energy System Operator (UK National Grid)	Accuracy	Completeness	Credibility	Latency	Plausibility

AWS Embodied Emissions Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
Constant hardware GWP values	Green Cloud Computing report, German Environment Agency	Accuracy	Completeness	Credibility	Latency	Plausibility	Waiting for German Environment Agency to confirm whether the report will be updated in future.	Published in 2021, hardware data potentially becoming outdated.
Variable hardware figures (units, capacity, weight, case type)	Boavizta database and AWS specifications	Accuracy	Completeness	Credibility	Latency	Plausibility	Add instance data through testing or available data sources.	Crowdsourced data backed up by CSP specifications.
CPU die size	Boavizta database	Accuracy	Completeness	Credibility	Latency	Plausibility	Build an independent dataset based on manufacturer data.	Crowdsourced data.
RAM density	TechInsights and Semiconductor Digest	Accuracy	Completeness	Credibility	Latency	Plausibility	Find more data.	Independent analysis of RAM hardware.
SSD density	Researcher analysis of hardware	Accuracy	Completeness	Credibility	Latency	Plausibility	Find more data.	Independent analysis of SSD hardware.
Variable instance figures (instance memory, storage, vCPU)	AWS instance specifications	Accuracy	Completeness	Credibility	Latency	Plausibility
Instance characteristics	Teads EC2 Instances Carbon Footprint Estimator	Accuracy	Completeness	Credibility	Latency	Plausibility		Teads Engineering carried out the first attempt at measuring cloud carbon in 2021. They now collaborate with Boavizta and their database forms the foundation of Boavizta’s data sets.
Network Storage Equipment	2017 GeekWire Article	Accuracy	Completeness	Credibility	Latency	Plausibility	Find updated information on CSP’s network storage servers.

Other AWS Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
Kubernetes hardware	ScaleFactory and Teads CPU Power Extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	Get confirmation from CSP about which CPUs are being used; find information on other hardware components.
Serverless hardware	ScaleFactory and Teads CPU Power Extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	Get confirmation from CSP about which CPUs are being used; find information on other hardware components.
Emissions Equivalent Miles Driven	DEFRA Conversion Factors, 2024	Accuracy	Completeness	Credibility	Latency	Plausibility
Emissions Equivalent Cups of Tea Brewed	Clever Carbon	Accuracy	Completeness	Credibility	Latency	Plausibility

Microsoft Azure

Updated: 14/01/25

Azure Operational Emissions Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
CPU Utilisation	Azure Monitor	Accuracy	Completeness	Credibility	Latency	Plausibility		Top N Chart only shows for top 5 virtual machines (VMs), but shows utilization per minute.
Billing & Usage	Azure Invoices and Command Line Interface	Accuracy	Completeness	Credibility	Latency	Plausibility
CPU Power	Teads CPU Power Extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	AMI testing to validate that Teads data is valid. Potentially run PowerAPI or other power draw monitor tools.	Some estimations and extrapolations were made for certain CPU power draws.
GPU Power	Teads GPU Power Draw or Telegraf	Accuracy	Completeness	Credibility	Latency	Plausibility	Needs GPU power sensing on user side.	Telegraf tracks in 1 minute intervals, creates new dashboard in Azure monitor.
RAM Power	Teads testing, plus 5% increment extrapolation	Accuracy	Completeness	Credibility	Latency	Plausibility	Lab testing to verify Teads’ RAM figures.	Memory usages can be tracked with Azure Monitor.
SSD Power	13 drives with capacities from 0.5TB to 8TB (See Appendix B)	Accuracy	Completeness	Credibility	Latency	Plausibility	Assess a wider range of SSD datasheets.
HDD Power	71 drives with capacities from 1TB to 24TB (See Appendix B)	Accuracy	Completeness	Credibility	Latency	Plausibility
Motherboard Power 42.5W	A1 Electronics	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a published study on motherboard consumption.	No demonstrable evidence behind A1 Electronics source, though its statement aligns with other anecdotal evidence.
NICs Power	NIC manufacturer datasheets	Accuracy	Completeness	Credibility	Latency	Plausibility		Power draw for MANA NIC is assumed based on similar models.
Network Storage Power	2017 GeekWire Article and assumption about 16kWh power draw	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a more recent source than GeekWire, and more specific to Azure.	Based on assumptions and 2017 source.
Intra- and Inter-region Networking	Ficher et al., 2021	Accuracy	Completeness	Credibility	Latency	Plausibility	Find an inter-region specific study.	Research paper only analyzed networking within one country.
External Networking	Collected research papers, 2014-2021 (See Appendix E)	Accuracy	Completeness	Credibility	Latency	Plausibility	Find a more recent study to validate these results.
Power Supply Efficiency Factor 1.1	2017 GeekWire Article and PSU manufacturers	Accuracy	Completeness	Credibility	Latency	Plausibility		2017 GeekWire article is outdated but manufacturer data provides credibility.
PUE 1.22	AWS 2024 blog post citing 2023 IDC report	Accuracy	Completeness	Credibility	Latency	Plausibility		1.2 figure is an average of the PUE of listed regions.
Grid Mix – Non-UK	Ember	Accuracy	Completeness	Credibility	Latency	Plausibility	Waiting for Ember to update their dataset to daily or hourly measurements.	1 month latency for every country. Unknown which datacenters generate their own power
Grid Mix – UK	National Energy System Operator (UK National Grid)	Accuracy	Completeness	Credibility	Latency	Plausibility

Azure Embodied Emissions Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
Constant hardware GWP values	Green Cloud Computing report, German Environment Agency	Accuracy	Completeness	Credibility	Latency	Plausibility	Waiting for German Environment Agency to confirm whether the report will be updated in future.	Published in 2021, hardware data potentially becoming outdated.
Variable hardware figures (units, capacity, weight, case type)	Boavizta database and Azure specifications	Accuracy	Completeness	Credibility	Latency	Plausibility	Add instance data through testing or available data sources.	Crowdsourced data backed up by CSP specifications.
CPU die size	Boavizta database	Accuracy	Completeness	Credibility	Latency	Plausibility	Build an independent dataset based on manufacturer data.	Crowdsourced data.
RAM density	TechInsights and Semiconductor Digest	Accuracy	Completeness	Credibility	Latency	Plausibility	Find more data.	Independent analysis of RAM hardware.
SSD density	Researcher analysis of hardware	Accuracy	Completeness	Credibility	Latency	Plausibility	Find more data.	Independent analysis of SSD hardware.
Variable instance figures (instance memory, storage, vCPU)	Azure instance specifications	Accuracy	Completeness	Credibility	Latency	Plausibility
Instance characteristics	Azure Invoices	Accuracy	Completeness	Credibility	Latency	Plausibility	Find out how Azure differ workload to CPUs and map which specific CPU is used for each VM.
Network Storage Equipment	Azure Files	Accuracy	Completeness	Credibility	Latency	Plausibility	Find updated information on CSP’s network storage servers.	Information is tangential and only covers data storage during transfer, not its entire storage time.

Other Azure Data Sources

Category	Source	Accuracy	Completeness	Credibility	Latency	Plausibility	Improvement	Notes
Kubernetes hardware	Microsoft resource on Cluster deployment & autoscaling	Accuracy	Completeness	Credibility	Latency	Plausibility	Further details on how the AutoScaler picks hardware and list of resources it chooses from.	Users can define a set of VMs for the node pool.
Serverless hardware	Microsoft resource on autoscaling and serverless	Accuracy	Completeness	Credibility	Latency	Plausibility	Further details on how the AutoScaler picks hardware and list of resources it chooses from.	Azure autoscaling.
Emissions Equivalent Miles Driven	DEFRA Conversion Factors, 2024	Accuracy	Completeness	Credibility	Latency	Plausibility
Emissions Equivalent Cups of Tea Brewed	Clever Carbon	Accuracy	Completeness	Credibility	Latency	Plausibility