furcate

Sovereign edge AI fabric · v0.1

Sovereign edge AI,
in motion.

The runtime, network, and compute fabric for AI at the edge — private, secure, by design. From a Raspberry Pi to a Jetson cluster to an air-gapped industrial gateway. The infrastructure layer powering the Ipnops family.

Built for builders shipping edge AI on real hardware.

edge nodes per cluster · KubeEdge

100k+

Wasm cold start · 100× containers

1-5 ms

Jetson Orin Nano Super · $249

67 TOPS

operation supported

air-gapped

01 · The shift

AI is leaving the cloud.

In 2026, AI inference doesn’t live in a hyperscaler’s datacentre by default anymore. It lives on a Jetson Orin Nano Super in a substation. On a Pi 5 with a Hailo-10H AI HAT+ 2 on a streetlight. On an ESP32-P4 in a sensor pod. On an air-gapped gateway in a defence contractor’s vault. Where the data is, where the privacy demands it, where the latency requires it.

Furcate is the substrate that makes that shift operable — runtimes, networking, fleet orchestration, hardware abstraction across consumer to industrial silicon. Sovereign, private, secure by design.

CLOUD92% cloud-onlyEDGE-CLOUD65% cloud-onlyEDGE-FIRST30% cloud-onlySOVEREIGN12% cloud-only
2018 · cloud-only2022 · hybrid edge2024 · edge-first2026 · sovereign edge

02 · Connect every layer

Not “an SDK.”
A complete edge stack.

Furcate composes the best open-source projects in edge AI into one coherent fabric — from the silicon up through the runtime, the network, the fleet orchestrator, and into the Ipnops intelligence platforms (or yours).

Edge silicon

NVIDIA Jetson family

Orin Nano Super · 67 TOPS · $249

Hailo-10H + Pi 5

40 TOPS INT4 · 2.5W · LLMs at 10 tok/s

Coral M.2 + USB

4 TOPS · 2W · TFLite-quantised

ESP32-P4 + Arduino + RP2350

TFLM · TinyML on microcontrollers

Runtimes & orchestration

TensorRT Edge-LLM · LiteRT · ONNX RT

FP8 · NVFP4 · INT4 quantization

KubeEdge · OpenYurt · K3s · Akri

100k nodes · 1M pods (KubeEdge benchmark)

WasmEdge · Wasmtime · WASI 0.3

1-5 ms cold start · 100× containers

Dapr · EdgeX Foundry · Eclipse Hono

Distributed runtime + protocol abstraction

Network, security, federated ML

Matter · Thread · LoRaWAN · private 5G + TSN

Mesh + sub-GHz + deterministic wireless

TPM 2.0 · TEE · attested boot

Hardware root of trust for every device

Sovereign by design

GDPR · HIPAA · FIPS · NIST · CMMC compliant

NVIDIA FLARE + Flower + OpenFL

Federated learning at production scale

Connect into Ipnops

Ipnops Coast

Coastal intelligence (cyan)

Ipnops Grid

Energy + physical systems (amber)

Ipnops Industry

Industrial automation (orange)

Bring your own platform

REST · gRPC · MQTT · OPC UA — protocol-agnostic

03 · The fabric

Every layer, composed.

Silicon at the bottom. Runtimes that turn weights into inference. An orchestrator that ships them safely to a fleet. A mesh that lets devices speak — to each other, to a sovereign control plane, to the customer’s perimeter and no further. Applications and Ipnops intelligence platforms at the top. Every layer composed of cited open-source projects; every interface explicit.

Application + Ipnops platformCoast · Grid · Industry · BYOMesh networkMatter · Thread · LoRa · 5G · TSNFleet orchestrationKubeEdge · OpenYurt · K3s · AkriDistributed app runtimeDapr · EdgeX · Hono · WasmEdgeAI runtimeTensorRT Edge-LLM · LiteRT · ONNX RTHardware + secure bootJetson · Pi · Hailo · Coral · ESP32 · TPM 2.0SOVEREIGN BOUNDARY

04 · Predict

Ask the fleet.

Interact with the fleet through natural language. The platform responds with real telemetry, real rollout simulations, and the lineage of every decision the agent makes — all without the data leaving your trust boundary.

furcate.ipnops.xyz / app / chat — live demo
● connected

You

live demo · queries cycle automatically · open the platform for the full experience

05 · Simulate & act

Test the rollout, before you ship it.

Discrete-event simulation for fleet operations — OTA campaigns, federated-learning rounds, model canaries, runtime swaps, mesh topology changes. The same simulator runs as the dispatch guard in production: every change is rehearsed before it ships to a single live device.

Scenarios

demo-fleet · Baseline OTA

● RolloutSim

DEVICES ON TARGET (%)T 12:00 · ON TARGET 0.0% · ROLLBACK 0.00%BASELINE · NO ROLLOUT CONTROLS
00:0012:0024:00

Closing

From someone else’s cloud,
to your hardware, your network, your terms.

As AI inference moves to the edge — onto the Jetsons in substations, the Pis on streetlights, the ESP32s in sensor pods, the air-gapped gateways in vaults — the world needs a substrate that runs intelligence at the customer’s perimeter, in the customer’s network, under the customer’s key, with the lineage of every model swap, every federated round, every dispatch decision auditable end-to-end.

Furcate is that substrate.