Energy procurement at data-centre scale
Continuous energy procurement at data-centre scale.
A hyperscale operator spends 30–50% of operating cost on electricity, and the cheapest electron at any given hour is rarely from the same supplier. Yet procurement is locked into long-dated bilateral contracts, because spot-market access is fragmented, onboarding a new supplier is slow, and verifying a carbon-intensity claim means chasing self-reported PDFs.
A Fabric-aligned procurement stack lets the operator — or an autonomous agent acting for it — treat energy as a continuous, signed, verifiable spot market. Suppliers publish kWh offers; the operator discovers and contracts hourly against policy; payment settles atomically against metered delivery; and every kWh is anchored for downstream carbon reporting.
What changes
Procurement moves from a quarterly contracting cycle to a continuous spot motion — the operator buys the cheapest compliant electron each hour.
Carbon-intensity claims are signed and replayable, not self-reported PDFs — so sustainability reporting is provable, not asserted.
Suppliers below the threshold for traditional broker access reach demand directly — widening the supply pool and sharpening price competition.
Variable-renewable supply finds price signal at the granularity it actually has — solar and wind get matched to load hour by hour.
An agent can run the loop — discovery, contracting, and settlement happen under explicit spend and carbon policy, with a full audit trail.
Where to start
Set up a sandbox Network Adapter, publish a single-supplier test catalogue, and run a simulated hourly procurement loop against it under a simple price-and-carbon policy. Migration to production is the same code with production endpoints — so validate the policy and the carbon-credential schema first.