How we run green — and how we measure it

1. Efficient by architecture

PeakSpitz's core runs as native, compiled code, not an interpreted application stack. This matters for energy because the work the processor does to serve a request is the work that gets metered — and emitted — for.

• • Independent basis: peer-reviewed studies ranking programming languages by energy use consistently place compiled languages among the most energy-efficient and interpreted ones among the least — often by an order of magnitude on equivalent tasks. We cite this as the category basis for our design choice. We do not derive a specific PeakSpitz multiplier from it.
• ✓ What we claim: PeakSpitz's engine is in the energy-efficient class of software by construction.
• ✗ What we don't: a numeric “X% less energy than competitor Y.”

2. Right-sized tenancy

Each customer runs an isolated, per-tenant deployment configured to that business — not a shared multi-purpose monolith with every module resident.

Unused capability isn't kept warm. Our architecture is configuration-driven rather than forked per vertical, so a tenant's running surface is scoped to what it actually uses. The result: you don't spend compute — or money — keeping features you don't use running.

3. Green infrastructure & region choice

PeakSpitz can run across multiple substrates and regions. Where geographically available, a customer can choose a greener deployment. The factual basis is the providers' own published data:

• • Datacenter efficiency (PUE): the operators we use publish Power Usage Effectiveness figures. Some EU operators we can run on are notable for low PUE achieved through techniques such as proprietary water-cooling. We link to the operator's current published figures rather than restate a number that can go stale.
• • Grid carbon intensity & carbon-free energy: regions differ widely in how clean their grid is. Where we run on a major public cloud, that provider publishes per-region carbon-free-energy percentages and grid carbon data; we use those to inform which regions count as “greener.”
• • EU data residency: greener-region options include EU-resident deployments.

We intentionally link to live provider data instead of quoting fixed PUE/CFE numbers here — quoting a frozen number is how methodology pages quietly become inaccurate. Ask us for the current figures for a specific region. We always keep the hedge: “where geographically available.”

Our roadmap centrepiece is customer-controlled efficiency:

• • Eco Windows: a customer will be able to define low-power windows on a daily-resolution schedule. During those windows the instance runs in a reduced-power mode, lowering its compute consumption — and therefore the customer's compute cost.
• • Shared-pool reclamation: capacity freed during a low-power window is returned to a shared pool rather than left idle, so it can do useful work elsewhere. The sustainability logic is fleet-level: reused capacity means the overall estate provisions and powers less total hardware for the same aggregate work.

Status & honesty: until this ships and we can describe the exact reclamation mechanism, we present it as a stated commitment, not a live feature. We'll update this section — with the real mechanism and any measurement — when it's available.

5. How we measure

We're explicit about our current measurement maturity rather than implying more than we have:

• • Today: we reason from structural facts (engine class, tenancy model) and provider-published infrastructure data (PUE, CFE, grid intensity).
• • Building toward: per-tenant compute/energy visibility, so a customer can see the footprint of their own usage — and the savings from Eco Windows — at daily resolution, the same granularity the schedule operates on.
• • Standards we look to: datacenter PUE/WUE reporting, cloud-provider carbon-free-energy methodology, and recognised software-carbon-intensity thinking. We align with these rather than inventing our own scoring.

6. What we do not claim (yet)

We will not say any of the following until each is measured and the method published:

When any of these becomes substantiable, it will appear here with its method, and only then on our sustainability page.

Changelog

• • 23 June 2026 — Page created (positioning posture; methodology baseline). Eco Windows & reclamation listed as roadmap.