Google Cloud Region Carbon Intensity — All 44 Regions (2024)
The same workload emits 226 times more carbon in one Google Cloud region than in another. Running it in Stockholm draws from a grid at 3 g CO₂e per kWh. Running it in Mumbai draws from one at 679. Nothing about the code changes — only the line on a deployment config.
This page publishes Google’s complete region carbon dataset as held in MasterBrain: all 44 regions, each with the annual average carbon intensity of its local grid and Google’s carbon-free energy percentage for that region in 2024.
Grid carbon intensity describes the electricity grid the data centre sits on, regardless of who is buying from it. That is the figure you use for location-based Scope 2. CFE% describes how much of Google’s hourly consumption in that region was matched with carbon-free generation — it is a statement about Google’s procurement, not about the grid. The two can diverge sharply: us-central1 sits on a 413 g CO₂e/kWh grid yet reports 87% carbon-free energy. Neither number alone tells you what to report, and the CFE figure is Google’s achievement rather than automatically yours.
The two metrics explained
| Grid carbon intensity | Google CFE% | |
|---|---|---|
| Describes | The local electricity grid | Google’s purchasing in that region |
| Unit | g CO₂e per kWh | Percentage of hourly consumption |
| Derived from | Hourly Electricity Maps data, aggregated to an annual average | Google’s hourly matching of consumption against carbon-free generation |
| Use it for | Location-based Scope 2 | Context on Google’s own progress; not a market-based factor for you |
| Changes if you move workload | Yes — different grid | Yes, but it is not your emission figure |
Google also flags a region as Low CO₂ when CFE% is at least 75% or grid intensity is at most 200 g CO₂e/kWh. 17 of the 44 regions currently qualify. That flag is carried in the tables below and in the cfe_qualifier field of the API response.
All 44 regions
Every Google Cloud region, cleanest grid first. Grid intensity is the 2024 annual average; CFE% is Google’s carbon-free energy share for the same year.
| # | Region | Location | Grid intensity (g CO₂e/kWh) |
Google CFE % | Low CO₂ |
|---|---|---|---|---|---|
| 1 | europe-north2 |
Stockholm, Sweden | 3 | 100 | ✓ |
| 2 | northamerica-northeast1 |
Montréal, Canada | 5 | 99 | ✓ |
| 3 | europe-west6 |
Zürich, Switzerland | 15 | 98 | ✓ |
| 4 | europe-west9 |
Paris, France | 16 | 96 | ✓ |
| 5 | europe-north1 |
Hamina, Finland | 39 | 98 | ✓ |
| 6 | northamerica-northeast2 |
Toronto, Canada | 59 | 84 | ✓ |
| 7 | southamerica-east1 |
São Paulo, Brazil | 67 | 88 | ✓ |
| 8 | us-west1 |
The Dalles, Oregon, USA | 79 | 87 | ✓ |
| 9 | europe-southwest1 |
Madrid, Spain | 89 | 87 | ✓ |
| 10 | europe-west1 |
St. Ghislain, Belgium | 103 | 84 | ✓ |
| 11 | europe-west2 |
London, UK | 106 | 79 | ✓ |
| 12 | us-west2 |
Los Angeles, California, USA | 169 | 63 | ✓ |
| 13 | europe-west12 |
Turin, Italy | 202 | 73 | — |
| 14 | europe-west8 |
Milan, Italy | 202 | 73 | — |
| 15 | europe-west4 |
Eemshaven, Netherlands | 209 | 83 | ✓ |
| 16 | southamerica-west1 |
Santiago, Chile | 238 | 92 | ✓ |
| 17 | europe-west10 |
Berlin, Germany | 276 | 68 | — |
| 18 | europe-west3 |
Frankfurt, Germany | 276 | 68 | — |
| 19 | asia-northeast2 |
Osaka, Japan | 296 | 46 | — |
| 20 | us-south1 |
Dallas, Texas, USA | 303 | 94 | ✓ |
| 21 | northamerica-south1 |
Querétaro, Mexico | 305 | 19 | — |
| 22 | us-east4 |
Ashburn, Northern Virginia, USA | 323 | 62 | — |
| 23 | us-east5 |
Columbus, Ohio, USA | 323 | 62 | — |
| 24 | us-east2 |
Lenoir, Georgia, USA | 340 | 42 | — |
| 25 | asia-northeast3 |
Seoul, South Korea | 357 | 37 | — |
| 26 | us-west4 |
Las Vegas, Nevada, USA | 357 | 64 | — |
| 27 | me-central1 |
Doha, Qatar | 366 | 1 | — |
| 28 | asia-southeast1 |
Jurong West, Singapore | 367 | 4 | — |
| 29 | us-central2 |
Council Bluffs, Iowa, USA | 372 | 88 | ✓ |
| 30 | me-central2 |
Dammam, Saudi Arabia | 382 | 1 | — |
| 31 | us-central1 |
Council Bluffs, Iowa, USA | 413 | 87 | ✓ |
| 32 | me-west1 |
Tel Aviv, Israel | 434 | 7 | — |
| 33 | asia-east1 |
Changhua County, Taiwan | 439 | 17 | — |
| 34 | asia-northeast1 |
Tokyo, Japan | 453 | 17 | — |
| 35 | australia-southeast2 |
Melbourne, Australia | 454 | 39 | — |
| 36 | australia-southeast1 |
Sydney, Australia | 498 | 34 | — |
| 37 | asia-east2 |
Hong Kong | 505 | 1 | — |
| 38 | asia-south2 |
Delhi, India | 532 | 29 | — |
| 39 | us-west3 |
Salt Lake City, Utah, USA | 555 | 33 | — |
| 40 | asia-southeast2 |
Jakarta, Indonesia | 561 | 18 | — |
| 41 | us-east1 |
Moncks Corner, South Carolina, USA | 576 | 31 | — |
| 42 | europe-central2 |
Warsaw, Poland | 643 | 40 | — |
| 43 | africa-south1 |
Johannesburg, South Africa | 657 | 15 | — |
| 44 | asia-south1 |
Mumbai, India | 679 | 9 | — |
By region group
The per-region figures aggregated by geography. The spread within a group is often wider than the gap between groups — Europe holds both the cleanest region on the estate and one of the dirtiest.
| Region group | Regions | Cleanest | Median | Dirtiest |
|---|---|---|---|---|
| North America (non-US) | 3 | 5 | 59 | 305 |
| Europe | 13 | 3 | 106 | 643 |
| South America | 2 | 67 | 152 | 238 |
| United States | 11 | 79 | 340 | 576 |
| Middle East | 3 | 366 | 382 | 434 |
| Asia | 9 | 296 | 453 | 679 |
| Australia | 2 | 454 | 476 | 498 |
| Africa | 1 | 657 | 657 | 657 |
Values in g CO₂e per kWh. Derived by GreenCalculus from the per-region figures above; Google does not publish region-group aggregates.
Application — worked example
Location-based Scope 2 for a cloud workload is consumption multiplied by the grid intensity of the region it runs in.
Scope 2 (location-based, kg CO2e) = kWh consumed x region grid intensity / 1000A workload consuming 100,000 kWh a year, placed in three different regions:
| Region | Location | Grid intensity | Annual Scope 2 |
|---|---|---|---|
europe-north2 |
Stockholm, Sweden | 3 | 0.3 t CO₂e |
us-central1 |
Council Bluffs, Iowa | 413 | 41.3 t CO₂e |
asia-south1 |
Mumbai, India | 679 | 67.9 t CO₂e |
Region choice alone moves this workload by 67.6 tonnes CO₂e a year — more than most efficiency programmes recover, achieved by changing one configuration value. Where latency, data residency and cost permit it, region selection is usually the single highest-leverage decision available on cloud emissions.
This covers electricity only. The embodied carbon of the underlying hardware, and Google’s own overhead beyond the machines running your workload, are separate questions this dataset does not answer.
Calculators that use these factors
The worked examples above are automated in these GreenCalculus tools, which read this dataset live from the MasterBrain so results always reflect the current vintage:
- Cloud storage carbon calculator — stored-data emissions by region.
- AI compute (training & inference) calculator — model compute located in a specific region.
- Data-centre PUE calculator — facility energy against regional grid intensity.
- Cloud compute allocation calculator — general cloud workload emissions.
Common reporting errors
- Reporting CFE% as your own achievement. The carbon-free percentage describes Google’s procurement in that region, not your contractual position. It does not make your market-based Scope 2 zero, and it is not a substitute for a supplier-specific emission factor backed by contractual instruments.
- Using the grid figure for market-based Scope 2. Grid intensity is a location-based factor by construction. A market-based figure depends on what you and your provider have contracted for, and must follow the GHG Protocol Scope 2 quality criteria.
- Treating an annual average as an hourly truth. These are annual aggregates of hourly data. A workload that runs only overnight, or only in summer, sits on a materially different grid mix than the annual mean implies.
- Assuming the figure is assured. Google publishes these as reported values, explicitly not third-party assured. That is acceptable for internal decisions and for a disclosed location-based estimate; it is weaker evidence than an assured supplier factor.
- Comparing across providers. Google, AWS and Azure use different methodologies, boundaries and vintages for their published regional figures. A cross-provider comparison built from each vendor’s own numbers is not like for like.
- Forgetting that regions move. Grids decarbonise and Google adds power purchase agreements. A region ranking taken in 2024 will not hold indefinitely, and a placement decision should be revisited when the dataset refreshes.
Methodology, boundaries & uncertainty
What the grid figure is. Google derives it from hourly grid mix data supplied by Electricity Maps, aggregated to an annual average for the balancing area the region sits in. It is a grid-average, location-based factor and carries the same interpretation as any other location-based Scope 2 factor.
What CFE% is. The share of Google’s electricity consumption in that region matched, hour by hour, with carbon-free generation. This is a stricter measure than annual renewable-energy-certificate matching, because it requires the generation to coincide with the consumption. It remains a statement about Google.
Not assured. The underlying basis field records these as reported values without third-party assurance. Where an assured figure is required — for a limited-assurance sustainability statement, say — treat this as supporting evidence rather than the primary source.
Vintage. Figures are the 2024 reporting year. Google refreshes them alongside its environmental reporting, and grid intensities move as generation mixes change. Apply the vintage contemporaneous with your reporting year rather than restating history on the newest edition.
Uncertainty. Grid-average intensities inherit the uncertainty of the underlying generation and interconnector data, which varies considerably by market — well constrained in Europe and North America, less so where hourly generation data is sparse. Differences of a few grams between neighbouring regions should not drive a decision; the differences that matter here are the order-of-magnitude ones.
Implementation & provenance chain
Every row traces to Google’s published region carbon table. Nothing is hand-transcribed: the tables are generated from the same MasterBrain rows the REST endpoint serves, and the generator asserts that all 44 source keys render exactly once before the page is built.
| Stage | What it is |
|---|---|
| Primary source | Google Cloud region carbon data, published at github.com/GoogleCloudPlatform/region-carbon-info and in Google Cloud documentation |
| Upstream grid data | Electricity Maps hourly grid mix, aggregated by Google to an annual regional average |
| Data layer | GreenCalculus MasterBrain — canonical keyspace digital.gcp.region.<region>, version-stamped |
| Filter | The feed is pinned to source id GCP_REGION_CARBON_2024. The wider digital section spans 25 sources, most of which are not redistributable — none of them appear here |
| Licence | CC BY 4.0 under the Google Cloud content licence; the same data is mirrored in an Apache-2.0 repository. Redistributable with attribution either way |
Data access — REST API & CSV
The full 44-row dataset is available as a machine-readable REST endpoint and as a flat CSV download. Both serve exactly the rows shown above, versioned by MasterBrain edition and citable.
cfe_percent, the Low-CO₂ qualifier, GHG scope/category and source references. Cache-Control: max-age=3600; X-GC-Version header signals dataset updates./wp-json/greencalculus/v1/gcp-region-factors
Click to generate ↓
Citation guidance
If you use this dataset in a published tool, report or academic work, cite Google’s primary source. The GreenCalculus compilation reference is optional but appreciated.
Google (2024). Carbon data for Google Cloud regions. Google Cloud Platform. Available at github.com/GoogleCloudPlatform/region-carbon-info. Licensed under CC BY 4.0. Grid carbon intensity derived from Electricity Maps hourly data.
Google Cloud region carbon primary citation
Cite this dataset (GreenCalculus compilation). A versioned, machine-readable snapshot of these factors is archived on Zenodo with a citable DOI:
Say, Jeremiah (2026). Google Cloud region grid carbon intensity & carbon-free energy percentage (2024) (machine-readable) (v2026.107). GreenCalculus. Zenodo. https://doi.org/10.5281/zenodo.21831161
GreenCalculus dataset DOI
Frequently asked questions
europe-north2 in Stockholm, at 3 g CO₂e per kWh with 100% carbon-free energy in 2024. Next are northamerica-northeast1 in Montreal at 5, then europe-west6 in Zurich at 15. At the other end, asia-south1 in Mumbai is 679 — a 226-fold spread across the estate. Rankings shift as grids decarbonise, so re-check when the dataset refreshes rather than treating a placement decision as permanent.
Grid carbon intensity describes the electricity grid the data centre sits on, whoever is buying from it. CFE% describes how much of Google’s own consumption in that region was matched, hour by hour, with carbon-free generation. They answer different questions and can diverge sharply: us-central1 sits on a 413 g CO₂e/kWh grid but reports 87% carbon-free energy, because Google has contracted heavily for clean generation in that market. For your location-based Scope 2 you use the grid figure.
No. CFE% is a statement about Google’s procurement, not about your contractual position. A market-based Scope 2 claim under the GHG Protocol requires contractual instruments that meet the Scope 2 quality criteria and are attributable to you. Google’s regional carbon-free share is useful context when choosing a provider or a region, and it belongs in narrative disclosure, but it does not on its own zero out your electricity emissions.
Location-based. Grid carbon intensity is a grid-average figure for the balancing area the region sits in, which is exactly what the location-based method calls for. A market-based figure has to reflect the contractual instruments you or your provider hold, and cannot be read off a regional grid average. Most organisations report both methods in a dual-reporting Scope 2 disclosure, and this dataset supplies the location-based half.
It is Google’s own designation, applied when a region reaches at least 75% carbon-free energy or a grid intensity of 200 g CO₂e per kWh or lower. Seventeen of the 44 regions currently qualify. Because the test passes on either condition, a region can carry the flag on a relatively dirty grid if Google’s procurement is strong there — so read the flag alongside the two underlying numbers rather than instead of them.
No, and the dataset says so explicitly in its basis field. Google publishes them as reported values derived from Electricity Maps hourly data. That is entirely adequate for internal placement decisions and for a disclosed location-based estimate, but it is weaker evidence than an assured supplier-specific factor. If you are preparing a statement subject to limited or reasonable assurance, treat these as supporting evidence and discuss the sourcing with your assurer.
Not directly. The three providers use different methodologies, boundaries and vintages for their published regional figures, so a table built by pasting each vendor’s own numbers together is not like for like. If you need a genuine cross-provider comparison, apply a single consistent grid dataset to each provider’s published region locations instead. That is more work, but it is the only version of the comparison that survives scrutiny.
No. These figures cover the electricity consumed, which is usually the largest share for a running workload but not the whole picture. The embodied carbon of the servers, storage and network equipment is a separate quantity, as is the data centre’s overhead beyond the machines doing your work. Region selection is the highest-leverage single decision available on the electricity half; it does not address the hardware half.
Version history
| Version | Date | MasterBrain | Summary |
|---|---|---|---|
| 1.0 | 2026-08-06 | v2026.105 | Initial publication. Complete Google Cloud region carbon dataset (44 regions, 2024): grid carbon intensity and CFE% per region, Low-CO₂ qualifier, derived region-group summary, location-versus-market-based guidance, worked placement example, 8-item FAQ, REST + CSV access. |