Explainer · 4 min read

What is AI
energy attestation?

Energy attestation is the act of signing a hardware-measured energy reading at the moment of capture and anchoring its hash on a public blockchain — so the number cannot be altered, forged or denied later.

Traditional energy reporting depends on spreadsheets and vendor dashboards. The numbers travel through human hands, are reconciled by hand, and are trusted by hand. An AI workload running on a GPU produces no native receipt; the operator writes one down.

Energy attestation replaces that paper trail with cryptography:

Why this matters now

Three regulatory waves arrive simultaneously:

What it is not

Energy attestation is not monitoring. Monitoring is a dashboard. Attestation is a signed record that can be verified by a stranger, in another jurisdiction, ten years from now — without any access to the original system.

It is not carbon accounting either. Carbon accounting takes an energy number and multiplies it by a grid factor. Attestation is upstream of that: it provides the energy number whose carbon claims will be derived.

See a real attestation Verify a signed, anchored certificate end-to-end — no API key.
Verify a certificate →

What gets signed

The canonical payload of an attestation includes:

energy_whHardware-measured value in watt-hours
duration_secondsSub-second resolution timestamps
measurement_sourcenvml / rapl / dual-source / manual
trust_model.cross_checksEvidence flags from independent sensors
trust_score.confidence0.0 — 1.0 graduated score
blockchain_anchorPolygon tx hash + Merkle root

Every field is part of the SHA-256 canonical hash. Changing any value invalidates the signature, which invalidates the Merkle leaf, which invalidates the on-chain anchor. The proof is structurally indivisible.