Sustainability | Dweve-reported AI Energy Comparison

Dweve-reported comparison of binary-first, fixed-point AI and an FP16 GPU baseline, with stated assumptions and measurement boundaries.

Dweve sustainability and AI energy

The sustainability page presents a Dweve-reported per-server comparison of a stated neurosymbolic workload with an FP16 GPU baseline. It describes binary-first compute, fixed-point arithmetic, standard hardware and air cooling, and lists the assumptions and external grid context alongside the figures.

The 96% and 366W figures are reported comparison values, not independent certification or a universal result for every workload and deployment.
Workload equivalence, correctness, hardware and software versions, power boundary and utilisation determine whether the comparison applies to a new measurement.

Choose the audience that matches your question

The page contains three selectable readings of the same subject.

For consumers

The sustainability page explains an energy comparison in plain language and shows why workload, hardware and measurement boundaries matter. Its figures are reported by Dweve for a stated comparison, not a universal promise for every AI deployment.

For businesses

Dweve reports a 96% lower-energy comparison for a stated neurosymbolic workload against an FP16 GPU baseline, alongside Dutch grid-capacity context. Treat the figures as a scoped comparison and review its assumptions before procurement.

For engineers

The reported comparison uses binary-first compute and correctly rounded fixed-point arithmetic on standard x86, ARM, RISC-V or FPGA hardware. The published 366W and 96% figures depend on workload equivalence, versions, power boundary and utilisation.