PrototypeEngineering in development

Designation

MF01

The first physical GRID AI Node.

GRID AI has existed as an architecture. MF01 is where that architecture becomes a machine. It is the prototype we are building, designed to test whether distributed AI compute behind an existing electrical connection can work technically, electrically, operationally and economically.

The thesis becomes a machine.
The machine creates the data.

  1. 01

    Thesis

    The distributed architecture.

  2. 02

    Machine

    MF01, in development.

  3. 03

    Measurement

    Instrumented electrical and workload data.

  4. 04

    Proof

    One Node before any network.

From assumption to infrastructure data

ConceptPrototypeModelAwaiting MF01 dataMeasured

Concept → prototype → model → measurement → validation → scale. Measured and validated are reserved for real MF01 data.

GRID·AIMF01
Power inHeat out

Conceptual industrial design — not final production hardware

The object

Infrastructure,
not a gadget.

MF01 is the first physical infrastructure prototype for GRID AI. It is not a gaming PC, a workstation, a consumer computer or a finished household appliance. It is intended to combine AI compute, power conversion and energy management, thermal handling, connectivity and remote control in one monitored enclosure.

Designation
MF01
Class
Prototype
Status
In development
Siting
Suitable existing connection
Storage in Node
None intended
Operation
Remotely monitored
What MF01 must answer01

One Node must work
before a network can.

01

Electrical

Can MF01 operate safely and reliably behind an existing electrical connection while measuring available capacity?

02

Behavioural

Can MF01 dynamically change its power consumption as available capacity changes while the property retains priority?

03

Thermal

Can the Node operate reliably within acceptable thermal conditions? Heat recovery may be measured separately but is not part of the base economic case.

04

Economic

Can the Node convert available electrical capacity into enough sellable compute value to support the infrastructure economics?

Objectives

Four measurable objectives. MF01 is built to answer them with data.

The MF01 proof program02

The build, in the open.

MF01 engineering is still being defined. The stages below represent what MF01 must ultimately prove. They are not promised dates and they are not completed milestones unless actual evidence exists.

  1. 01

    Bench build

    Assemble and validate the initial MF01 hardware architecture.

    Planned
  2. 02

    Behind the connection test

    Connect MF01 to a representative electrical environment and establish reliable measurement of Node consumption and property demand.

    Planned
  3. 03

    Property priority test

    Demonstrate that MF01 can reduce its power consumption when the property requires more capacity.

    Planned
  4. 04

    Compute workload test

    Run real AI compute workloads and observe performance under normal and dynamically constrained power conditions.

    Planned
  5. 05

    Remote operations test

    Operate, monitor, control, restart and diagnose MF01 remotely.

    Planned
  6. 06

    Economic validation

    Connect actual compute activity, electricity consumption, availability and operating cost to measured Node economics.

    Planned
Live measurements03

Every figure here becomes
measured, or it does not stand.

Continuous Node draw

—

Awaiting MF01 dataWhat electrical power does MF01 actually consume during different workload states?

Electrical availability

—

Awaiting MF01 dataHow much usable capacity is actually available to MF01 behind the property connection over time?

Dynamic power response

—

Awaiting MF01 dataHow quickly and reliably can MF01 reduce or increase its consumption when available electrical capacity changes?

Measured compute rate

—

Awaiting MF01 dataWhat revenue per actual sellable compute unit or GPU hour can GRID AI measure from real workloads? The model value is an underwriting assumption until MF01 measures it.

Effective compute utilization

—

Awaiting MF01 dataHow much of MF01's technically available compute capacity becomes actual revenue producing compute?

Technical uptime

—

Awaiting MF01 dataHow much of the measurement period is MF01 technically available to perform compute?

Node electricity cost

—

Awaiting MF01 dataWhat is the actual electricity cost attributable specifically to MF01?

Node contribution

—

Awaiting MF01 dataWhat economic contribution remains after the Node level costs included in the approved economic methodology?
See how MF01 data feeds the economic model →
Secondary measurements04

Measured because they may matter later.

Heat, noise, network behaviour and reliability events are recorded as secondary data. They do not determine the base GRID AI business case, but they may create future operational or economic value.

Thermal performance

Awaiting MF01 data

Noise

Awaiting MF01 data

Network performance

Awaiting MF01 data

Maintenance events

Awaiting MF01 data

Faults

Awaiting MF01 data

Workload interruptions

Awaiting MF01 data

Recoverable heat fraction

Awaiting MF01 data
Economic relationship05

A Node cannot create value
merely by existing.

Conceptual economic relationship

Compute demand×Electrical availability×Technical uptime
↓ Effective sellable compute↓ Measured compute revenue↓ Node economics

There must be paying compute demand. The Node must have electrical capacity available. The Node must be technically operational. And GRID AI must convert that capacity into actual paid compute. This is what MF01 must begin to measure.

Property priority06

Property first.
Node adapts.

MF01 must test whether Node power can follow available capacity without requiring the property to adapt its normal electricity consumption to the Node. The property keeps priority; the Node reduces its draw when capacity is needed elsewhere.

Awaiting MF01 data
Model versus measured07

From model to
measured.

Each pair holds the current model value and the field designed to replace it with real infrastructure data once MF01 operates.

Compute rate

Model

€0.39 / rented GPU hour ($0.45)

Measured

Awaiting MF01 data

Utilization

Model

75%

Measured

Awaiting MF01 data

Node electricity

Model

€0.28 / kWh

Measured

Awaiting MF01 data

Node contribution

Model

approx. €3.418 / year

Measured

Awaiting MF01 data

Model values are illustrative underwriting assumptions. Underwriting assumptions are model inputs and are not live market prices. Current external GPU market observations are shown only in the live compute market section on the homepage.

What happens after MF01?08

Proof before scale.

If the case works

Technical and economic case holds

↓

Build 10

↓

Test repeatability

If the case does not work

Change the architecture

↓

Test again

BUILD 10 is the intended next proof stage if MF01 produces sufficient evidence. It is not a committed deployment.

See the path from proof to scale →
MF01 disclosure09

MF01 disclosure

MF01 is a prototype in development. Technical architecture may change during engineering. Tests shown are the intended proof program. Measured fields remain empty until real operating data exists.