furcateDeploy ──►
Furcate · siteSite servers & small data centres

FURCATE
SITE

Deploy Furcate to server-class hardware on the operator's perimeter — one box in a comms cabinet or a three-node HA rack in the plant room. Confidential computing, attested end to end. This is where sovereign AI infrastructure actually lives: a data centre the size of the problem, on power you already have.

Furcate Site · reference
Furcate Site

Specification

HardwareDell PowerEdge XR · HPE ProLiant DL145 Gen11 · Supermicro IoT SuperServer · Lenovo ThinkEdge SE360 / SE450 · ASRock 1U
RAM64 GB – 1 TB · GPU optional, attested when present
DeployCLI · desktop
OSLinux · K3s 3-node HA control plane
ConfidentialityIntel TDX · AMD SEV-SNP · NVIDIA Confidential Computing for GPU-attested workloads. Measured boot to signed workload, provable on the perimeter.
RoleThe site brain. Runs the heaviest models, local training and federation, the orchestrator, the immutable local ledger. The Furcate Site Server class — a small data centre that makes the distant one optional.
FootprintKilowatts, not megawatts. Standard rack power and cooling — no new substation, no grid queue, no gigafactory.
NetworkWired Ethernet backbone of the site mesh. Serves model-to-model requests from Micro, Small and Base neighbours; gates cloud-burst and federation by explicit grant.
IdentityIssued a Furcate machine identity — a local DID — at first deploy.

Site server class · the data centre, right-sized