Skip to content

Carbon Intensity

Grid carbon intensity (the amount of CO₂e emitted per kilowatt-hour of electricity) is the bridge between energy consumption and emissions. Ambra Spectrum uses hourly, location-based intensity values matched to each cloud region.

Primary source: Nodera Intensity API

The Nodera Intensity API provides hourly carbon intensity values (gCO₂e/kWh) for every supported cloud region. Each Spectrum data point is matched to the intensity value for its region and the hour containing its timestamp.

Hourly granularity captures the real variation in grid carbon intensity throughout the day: renewables peak during daytime. Intraday variation can reach up to 10× in grids with high renewable penetration, meaning the same workload can produce vastly different emissions depending on when it runs. Annual, monthly and daily averages mask this variation entirely.

Estimates produced using the Nodera Intensity API are tagged with intensity_source = "nodera_api".

Static fallback

When the Nodera Intensity API is unavailable, Spectrum falls back to static annual average intensity values per region. These are derived from:

  • US regions: EPA eGRID (Emissions & Generation Resource Integrated Database)
  • International regions: Ember Energy yearly electricity data
  • GCP regions: Google Cloud Platform region-carbon-info repository

These provide a reasonable baseline but do not capture hourly variation. Estimates produced using the static fallback are tagged with intensity_source = "static_fallback", enabling clients to distinguish them in reporting.

A global default is applied for regions not covered by either source.

Intensity matching

For each carbon estimate data point:

  1. The cloud region is mapped to its underlying datacenters' physical addresses
  2. From each datacenter address, Nodera identifies the grid connection point, the specific node on the electricity transmission network where the datacenter draws power
  3. Nodera computes local carbon intensity at each grid connection point, then averages across all connection points serving the cloud region
  4. The data point's timestamp is floored to the containing hour, and the hourly intensity value for that (region, hour) pair is retrieved

Because intensity is resolved at the grid connection level, Spectrum captures geographic variations within a single country. For example, a datacenter in northern France connected to a nuclear-heavy grid segment will show a different intensity than one in the south relying more on gas peakers. This is fundamentally different from a national-average approach.

This also ensures that a 5-minute data point at 14:23 uses the 14:00–15:00 intensity value, not a daily or monthly average.

Location-based vs market-based

By design Spectrum currently provides location-based intensity only to provide a true carbon intensity based on grid physics. This reflects the average carbon intensity of the running data centers based on their location on the regional electricity grid. In line with Nodera’s commitment to deliver true carbon intensity values based on grid physics it does not account for renewable energy certificates (RECs), power purchase agreements (PPAs), or carbon-free energy (CFE) commitments.