GPU Fleet

Roadmap context

These are operator-owned development and cook rigs — the real sovereign-compute fleet the house runs on. The edge appliance program and any customer-facing edge fleet service built on top of it are roadmap, not a shipped offering. The hardware below is real; the productized edge service is not yet live.

This is the owned sovereign-compute fleet that backs the roadmap — the rigs the work actually runs on:

• swarmrails — Xeon w9-3475X · 256 GB RAM · RTX PRO 4500 (32 GB) + RTX PRO 6000 (96 GB)
• whale — Ryzen 9 5900X · 64 GB RAM · RTX 3090 (24 GB)
• signal-edge-01 — Jetson Orin Nano (8 GB)
• zima-edge-1 — Intel N150 (14 GB)

Combined VRAM: 160 GB (32 + 96 + 24 + 8 = 160 GB across the four GPUs/accelerators) · Combined RAM: 342 GB (256 + 64 + ~22 across the boxes). The VRAM total checks out internally.

---

🐝 Operator-grade · books and records · to the shed.