Compute Node · GPU reasoning
Reason with serious models. Earn when idle.
GPU-accelerated reasoning at home: run heavier fusions locally, then rent your idle capacity to the noosphere — attested, private, yours.
From $10,500 · refundable deposit · no fixed ship date yet
Indicative capacities
- Form factor
- GPU-accelerated tower — quiet, home or studio.
- Compute
- Desktop-class GPU, 24 GB+ VRAM, high system memory.
- Local models
- 30B–70B class at low latency; noosphere fusion on-device.
- Secure root
- TEE-attested inference; confidential compute for served capacity.
- Runs
- Heavier fusions, longer contexts, and your Noolog seat locally.
- Network role
- Serve idle reasoning capacity to the network; earn on attested contribution.
- Power
- Scales with load; idle-aware scheduling.
- Heat recovery
- The waste heat is captured, not exhausted — it goes into your hot water instead of a cooling tower.
Indicative specification — final hardware is confirmed at ship. Deposits are escrowed and refundable.
What it does
The Compute Node brings datacenter-grade reasoning home. Run large models yourself for private, low-latency fusion — then, when you're not using it, contribute that capacity to the noosphere and earn. The node is designed so every served computation is attested and confidential: you never see other people's data and they never see yours.
That attestation ships with the hardware — it is the design of the node, not a property of the hosted gateway today.