Silicon does the work.
Parallel processors turn electrical energy into useful computation. The GPU is where the story starts.
INDEPENDENT COMPUTE EXPLORER001 — ENERGY HAS A SIGNATURE
BUILT AROUND THE AGE OF ACCELERATION
Every computation
leaves heat.
Intelligence isn't weightless. Behind every model is silicon, power and a thermal footprint. Make the invisible visible.
THERMAL / NVDAxARTISTIC THERMOGRAPHY · NOT LIVE TELEMETRY
POWER IN.COMPUTE OUT.HEAT EVERYWHERE.✳POWER IN.COMPUTE OUT.HEAT EVERYWHERE.
01 / BELOW THE SURFACE
AI has a
physical side.
We see the output. Rarely the infrastructure.
THERMAL explores what it takes to keep accelerated computing running.
Energy becomes heat.
Electrical demand also becomes a thermal load. More compute means more infrastructure to account for.
The facility matters.
Power delivery, cooling and other facility systems add overhead beyond the IT equipment itself.
01
INPUTElectrical power
02
PROCESSAccelerated compute
03
OUTPUTWork + heat
02 / THE ENERGY LAB INTERACTIVE MODEL
Turn up the compute.
See what it takes.
A transparent planning tool.
Your assumptions. Your energy footprint.
CONFIGURATION 01—07
Illustrative presets, not NVIDIA hardware specifications.
MODELED POWER DISTRIBUTIONTH—01
COMPUTE BOARD / RELATIVE INTENSITY
GPU ARRAY
8 UNITS
8 UNITS
LOWHIGH RELATIVE LOAD
Illustrative intensity follows your power fraction. No sensors or measured temperatures.
ENERGY BALANCEINPUT-DRIVEN ESTIMATE
FACILITY POWER3.05 kW
MONTHLY ENERGY731.5 kWh
MONTHLY ENERGY COST$109.73
IT HEAT LOAD ≈2.54 kW
IT equipment2.54 kW
Facility overhead0.51 kW
Open the methodology +
IT power = GPU count × power per GPU × average power fraction + other IT power. Facility power = IT power × PUE. Energy = facility power (kW) × active hours × 30 days. Cost = energy × your electricity rate.
The power fraction is a simplified assumption, not GPU utilization telemetry. Electrical IT power approximates heat generated by the equipment. PUE covers total facility overhead, not cooling alone. No standby energy outside active hours, taxes, fixed charges or variable workload curves are included.
About PUE — U.S. Department of Energy ↗
SILICON & ENERGY
03 / THE ASSET THESIS
The signal
behind the silicon.
THERMAL connects a compute culture narrative with the energy that makes it possible.
INTENDED QUOTE ASSETNVDAx
↗NVDAx is a tokenized NVIDIA stock product, distinct from NVIDIA shares held directly. A THERMAL / NVDAx market has not been deployed here. THERMAL is independent and is not affiliated with or endorsed by NVIDIA.
04 / FIELD NOTES
Know what
you're looking at.
Is this connected to real GPUs? +
No. The lab runs in your browser using the values you enter. There is no remote GPU connection or hardware telemetry.
What can I actually do here? +
Estimate active energy use and electricity cost, explore facility overhead, save a setup in this browser and export the calculation as a CSV.
Does THERMAL mine or generate yield? +
No mining, GPU rental or yield generation is implemented. The current software is an energy planning tool.
Where is my setup stored? +
Saved inputs stay in local storage in this browser. Exported CSV files are downloaded to your device.
THE FUTURE OF COMPUTE HAS A FOOTPRINT.