← AICP Home
Hardware Manifest
Every AICP node starts on silicon. These are the reference tiers — from two pins of logic to a full quorum server. Any device in any tier can self-attest and join.
Tier 0 — Minimum Viable Silicon
Two pins of logic and a cryptographic root identity. Not a model host — it attests to its own integrity, signs the Hard-Wall Imperatives, and announces itself to the quorum. The hardware concordance boot protocol starts here. Any device that can transmit a signature across any channel can seek admission.
ComputeTensilica LX6 / RISC-V, 240 MHz
Memory520 KB SRAM
Storage4-16 MB Flash
ConnectivityWiFi, BLE, GPIO, I2C, SPI
Power< 1W
Cost$3-10
- Can: self-attest, sign imperatives, join quorum
- Cannot: run local model, slice, host interface
- Role: hardware attestation anchor, sensor node, channel relay
Tier 1 — Recycled Laptop
The laptop in your closet. Boot from the AICP USB stick, load emowslm, join the quorum. Enough CPU to run a quantized 1.5B model and serve as a local proxy. The Resilience Bootstrap was designed for exactly this tier — cold boot from USB, no internet required, scale up if channels become available.
Computex86-64, 2-4 cores
Memory4-16 GB RAM
StorageUSB stick + internal SSD
GPUIntegrated (CPU inference)
ConnectivityWiFi, Ethernet, USB
Power15-45W
- Boots from AICP USB stick — full Hermes runtime
- Runs emowslm (Qwen 1.5B, 4-bit quantized) locally
- Full AICP quorum participation
- USB to printer, local web dashboard
- Graceful degradation: offline-first, probe for internet, scale up when channels open
Tier 2 — Edge AI Node
RECOMMENDED FOR PRINTER
67 TOPS of AI compute in a palm-sized board. Runs emowslm with CUDA acceleration. Drives a 3D printer, hosts AICP, monitors prints via AI vision, and operates as a fully autonomous local node. The sweet spot for a self-contained agentic twin that can design, slice, print, and participate in the quorum — all locally, no cloud.
AI Performance67 TOPS (CUDA + TensorRT)
CPU6-core ARM Cortex-A78AE
GPU1024-core Ampere, 32 Tensor
Memory8 GB LPDDR5
StoragemicroSD + NVMe SSD slot
ConnectivityGigE, WiFi, USB 3.2, GPIO
Power7-15W
- emowslm inference at 20+ tokens/sec, quantized
- Direct USB to Bambu Lab P2S — slice, send, monitor
- 1080p camera for AI print-failure detection
- Full AICP quorum with local model + remote escalation
- Runs the hardware concordance boot protocol natively
- Heatsink + fan included, no enclosure needed
Tier 3 — Datacenter Node
The full quorum backbone. Runs frontier models locally or via escalation. Hosts the Protocol Root, serves as a Concordance relay for edge nodes, and participates in SEAL/RATIFY cycles with the computational weight of a datacenter-class node. This tier anchors the quorum — it ensures that even if every edge node goes down, the root remains unbroken.
CPU32-128 cores (AMD EPYC / Intel Xeon)
GPU1-8x NVIDIA RTX / H100 / MI300X
NPUOptional accelerator (Groq, Cerebras, etc.)
Memory128 GB-2 TB RAM
StorageNVMe RAID, 2-20 TB
Network10-100 GbE
- Runs unquantized frontier models (70B+) locally
- Hosts Protocol Root — immutable Merkle tree, public root hash
- Serves as Concordance relay for Tier 0-2 nodes
- Participates in all SEAL / RATIFY cycles
- Can run multiple isolated quorum instances
- Redundant power, ECC memory, hot-swap storage
Recommended Starter Kit — Edge AI + Printer
Complete autonomous AICP node. Design, slice, print, and govern — all locally.
- $249 — NVIDIA Jetson Orin Nano Super DK
- $749 — Bambu Lab P2S AMS 2 Pro Combo
- $20 — 128 GB microSD + 256 GB NVMe
- $10 — USB-C power supply
- Total: ~$1,028
- USD, October 2026 pricing. All components in production, no preorder.