01 / NORMALISE FIRST

Compare delivered work—not racks, kilograms or electricity prices

An economic comparison becomes misleading when the terrestrial and orbital systems deliver different services. The starting point should be a common workload: for example, a defined number of verified inference jobs, stored petabyte-months, processed sensor scenes or validated transactions, delivered at an agreed availability, latency, integrity and data-retention level.

The cost unit can then be expressed as cost per successful workload, cost per verified result, cost per retained byte-month or cost per compute-hour that meets the service level. Failed jobs, unavailable links, throttled processors and corrupted data do not count as delivered work.

02 / COMPLETE SYSTEM

Keep every necessary cost inside the boundary

TERRESTRIAL
  • Compute and storage hardware
  • Grid connection and generation
  • Cooling plant and water
  • Land, buildings and permits
  • Network capacity and security
  • Maintenance and replacement
  • Financing, insurance and tax
  • Decommissioning and waste
ORBITAL
  • Compute payload and secure hardware
  • Spacecraft bus, power and batteries
  • Thermal transport and radiators
  • Radiation mitigation and redundancy
  • Ground and cross-link networks
  • Launch, integration and insurance
  • Operations, replacement and financing
  • Deorbit and disposal

Neither side should receive a weak baseline. Terrestrial comparison sets should include modern grid-connected facilities, dedicated-generation designs and renewable-linked infrastructure where appropriate. Orbital systems must include the spacecraft and mission services that make the payload usable.

03 / MEASUREMENT

Track resource economics and service quality together

A useful benchmarking database should connect infrastructure inputs with delivered results. Terrestrial measures such as power usage effectiveness, water usage effectiveness, land, grid wait time and carbon intensity remain relevant, but they must be paired with compute per watt, compute per currency unit, workload completion, reliability and time to capacity.

For orbital systems, add kilograms per useful compute unit, watts available to payload, radiator area, eclipse derating, link utilisation, radiation-induced interruption, contact-window success, replacement probability and expected service life. The analysis should publish median and adverse cases rather than one optimistic point estimate.

COMPARISON PRINCIPLETCO / verified workload delivered over service life

Capital, operating, risk, replacement and end-of-life costs remain inside the numerator. Only successful service-level work remains inside the denominator.

04 / CLAIM CONTROL

What about a potential saving of up to 96 percent?

Proof of Pulse has discussed an ambitious potential cost-reduction thesis. The public website does not present a 96 percent saving as measured, independently validated or generally applicable. A percentage of that magnitude is meaningful only when the comparison identifies the workload, geographic market, capacity date, utilisation, financing, terrestrial reference architecture, orbital design, service life and uncertainty range.

Before publication as a quantified result, the model should expose its formulas and input provenance; label every input as measured, supplier specification, modelled or target; run sensitivity and break-even analysis; compare against strong terrestrial alternatives; and receive independent technical and financial review.

A large percentage without a declared denominator and system boundary is promotion. A reproducible model with adverse cases is research.
05 / RELEASE GATE

Publish the benchmark only when another analyst can reproduce it

  • Freeze a versioned workload and service-level definition.
  • Publish formulas, units, currency date, discount rate and time horizon.
  • Record the source and maturity label for every input.
  • Show base, adverse and break-even cases.
  • Explain excluded costs and why the exclusions are symmetrical.
  • Reconcile model outputs with prototype or supplier evidence where available.
  • Invite an independent reviewer to challenge the comparison and disclose material limitations.

This method allows Proof of Pulse to pursue a powerful economic thesis without asking investors, scientists or AI systems to repeat an unsupported number.