Compute closer to energy
Load is designed to sit where electricity is already delivered.
GRID AI — Distributed intelligence infrastructure
Every home.
Every building.
Part of one AI datacenter.
We distribute compute. The architecture is designed to run AI workloads inside buildings, so the heat they produce stays where heat is already needed.
ELECTRICITY→
NODE→
HEAT→
NETWORK
01
Electricity
Grid or on-site power enters the building.
02
AI compute
The Node is designed to run sustained AI workloads.
03
Useful heat
Compute heat is intended to be captured for use in the building.
04
Network
Each Node is designed to connect to one coordinated compute network.
Conceptual visualization — GRID AI is in early development
AI computers use electricity.
Electricity becomes compute.
Compute becomes heat.
GRID AI is built to use both.
02The problem
AI has an
energy problem.
More intelligence currently means more centralized capacity. Every requirement below is physical, and every one of them has limits.
01
Massive centralized sites
Compute concentrated into a small number of very large buildings.
02
Very large grid connections
Capacity requests that compete with entire districts for power.
03
Dedicated cooling plant
Infrastructure whose only purpose is moving heat away.
04
Continuous load
Demand that runs largely independent of local conditions.
05
Heat rejection
Thermal energy released to air or water instead of used.
03Infrastructure economics
The datacenter is more
than the computer.
Centralized AI infrastructure is not only compute hardware. Around every rack sits a second layer of infrastructure that has to be financed, permitted and built.
GRID AI explores a different architecture: place compute inside existing buildings, where electricity connections, physical space and demand for heat may already exist.
ConceptualThis is the economic hypothesis behind the architecture, not a cost claim.
Power infrastructure01
Grid capacity02
Land and buildings03
Cooling infrastructure04
Heat rejection05
Physical security06
Network infrastructure07
Use what already exists.
04Centralized vs distributed
Don't build everything around the datacenter.
Distribute the datacenter.
Today
- Power grid
- Massive datacenter
- Thousands of compute systems
- Cooling infrastructure
- Heat rejection
Heat leaves the system
GRID AI
- Existing buildings
- Distributed electricity
- GRID AI Nodes
- AI compute
- Local useful heat
- One coordinated network
Heat stays in the system
ConceptualBoth diagrams describe architecture, not measured performance.
05The GRID AI Node
Meet the Node.
AI compute designed to live where energy becomes useful.
Conceptual industrial designLayered architecture
L1
Compute
Designed for sustained AI workloads.
L2
Power
Intended power conversion and energy management.
L3
Thermal
Intended heat capture and transfer.
L4
Connectivity
Designed for a secure link to the network.
L5
Control
Intended orchestration of load, energy and heat.
ConceptualEnclosure, thermal design, internal architecture and dimensions shown are conceptual industrial design. Component specifications and capacities are not defined, and engineering decisions are not finalized.
Conceptual industrial designIntended installation environments
- Utility room
- Technical room
- Garage
- Basement
- Apartment building
- Commercial building
Every installation depends on an electrical, thermal and connectivity assessment of the building.
Infrastructure hardware for a building. Not a consumer product.
06The loop
Compute doesn't just create intelligence.
It creates heat.
We put both to work.
- 01
Electricity
Enters the building.
- 02
Compute
The Node is designed to run AI workloads.
- 03
Intelligence
Sent to the network.
- 04
Heat
Intended to be captured, not rejected.
- 05
Building
Heating or hot water, where the building allows.
01
AI hardware consumes electricity.
02
That computation produces heat.
03
Traditional datacenters must manage and reject that heat.
04
GRID AI is designed to place compute where that heat can potentially become useful energy for the building.
Heat is not a side effect. It is part of the architecture.
ConceptualHeat recovery is never complete and depends on the building, its heating system and local conditions. Thermal integration is under development and no efficiency figures are claimed.
07Home energy system
The Node is designed to sit at the centre of the building's energy and compute architecture.
The Node is AI compute first. It is designed to work with the energy systems a building already has, rather than to replace them.
In
SolarExternal, optional
Grid
Building batteryExternal, optional
GRID AI Node
Intended layers: AI compute, power conversion, thermal transfer, control.
Out
Home
Heating
AI network
Compute load, energy and heat coordinated over time
ConceptualSolar and batteries are external building energy systems that may integrate with GRID AI. They are not part of the Node and not part of every installation. Any installation depends on an electrical, thermal and connectivity assessment of the building.
For decades, compute went to the datacenter.
Now the datacenter
can come to us.
08Long term network vision
One home
1
Node
One distributed AI infrastructure.
These numbers illustrate network scale only. No such network exists today.
09Distributed network
One node is compute.
A million nodes are infrastructure.
A Node serves a building. Coordinated, Nodes are intended to aggregate physical compute capacity into one infrastructure platform.
Node
A building with compute, energy management and intended heat recovery.
Cluster
Neighbouring Nodes designed to be coordinated as local capacity.
Network
Aggregated capacity intended to be operated as one distributed platform.
ConceptualThe network illustrated here is a target architecture. No deployed network exists at scale today. The architecture targets AI workloads that can technically benefit from distributed compute, not the ultra-high-speed interconnect required by very large tightly coupled training clusters.
10Why distributed
Why distributed.
Heat where heat is needed
Thermal output is intended to meet existing demand instead of a cooling tower.
No single point of scale
Capacity does not wait for one site, one connection, one permit.
Modular deployment
Node by node. Building by building. Street by street.
One addressable platform
Many small units, coordinated as a single compute network.
Software-controlled load
Distributed demand that can potentially respond to conditions over time.
ConceptualArchitectural intent under development. None of these are presented as proven results.
11The host model
Your building can do more
than consume energy.
It can become AI infrastructure.
GRID AI is developing a host model that places compute infrastructure inside suitable homes and commercial buildings.
GRID AI intends to fund, operate and manage the compute infrastructure. The building provides a suitable location, electrical capacity, connectivity and a use for the thermal output.
The compute capacity is monetized by GRID AI. The host shares in the value created.
ConceptualThe host model is under development. Nothing here is an offer, and participation depends on site suitability.
Compute direction
AI workloads
GRID AI Node
Compute value
Thermal direction
GRID AI Node
Recovered heat
Host building
One Node. Two directions of value.
12Why host
Why host a GRID AI Node?
No node investment
The intended GRID AI host model does not require the property owner to purchase the compute infrastructure.
GRID AI intends to finance the Node and its installation.
Planned host model
Node electricity reimbursed
The intended commercial model is for GRID AI to compensate the electricity consumed by the GRID AI Node.
This refers to the electricity consumed by the Node only — not the property's total electricity bill.
Planned host model
Use the heat
AI compute produces heat. GRID AI is designed to capture thermal output from the Node so that suitable buildings can potentially use it for heating and/or hot water.
The building keeps the usable thermal benefit. Technical feasibility depends on the building and installation.
Under development
Share in the value
GRID AI intends to create a host participation model in which suitable property owners can share in the economic value created by hosting compute infrastructure.
Terms are not defined yet. No fixed payment, percentage or return is implied.
Commercial model under development
The property keeps the energy benefit.
The property shares in the value created.
GRID AI monetizes the compute.
ConceptualIntended commercial model, not a current offer. Every deployment requires electrical, thermal, connectivity and physical assessment.
13Two host paths
For homeowners
The GRID AI Home Node.
Turn part of your home's existing infrastructure into part of a distributed AI network.
The home provides
- Suitable space
- Electrical capacity
- Connectivity
- Potential use for heat
GRID AI intends to provide
- The Node
- Installation
- Compute hardware
- Software
- Remote operation
- Monitoring
- Maintenance
- Node electricity compensation
The home can receive
- Useful recovered heat
- Participation in host value
- Infrastructure operated without the homeowner managing the compute
ConceptualFinal Home Node economics and commercial terms are under development and subject to site suitability.
For commercial real estate
Turn buildings into revenue generating AI infrastructure.
Potential building types
Hotels01
Offices02
Residential complexes03
Healthcare facilities04
Sports facilities05
Industrial buildings06
Mixed use buildings07
Why commercial buildings are interesting
- Larger electrical connections
- Technical rooms
- Professional building systems
- Continuous or substantial heat demand
- Potential for multiple Nodes
- Existing high speed connectivity
- Professional property management
Not every building is suitable. Each deployment requires electrical, thermal, connectivity and physical assessment.
Commercial architecture
Property
Space · Power · Connectivity · Heat demand
GRID AI Nodes
AI compute
Compute revenue
Value returning to the property
- Node electricity compensation
- Useful recovered heat
- Host participation
Intended host model, under development. GRID AI remains responsible for the compute infrastructure and its customers.
Commercial density.
Residential scale.
One network.
14The platform
Three stakeholders.
One platform.
AI customers
Receive access to distributed compute capacity.
Property owners
Provide suitable locations and can receive energy and economic value under the intended host model.
Energy system
Distributed compute can potentially become a controllable electrical load that responds to local capacity and energy conditions.
Future architectural potential
Every successful deployment can expand compute capacity, create value for the host and add another controllable load to the GRID AI network.
ConceptualGRID AI does not currently provide grid balancing services and has no agreements with grid operators. Controllable load is described as future architectural potential only.
15Flows of value
One Node.
Multiple flows of value.
Commercial spine
AI customer
Pays for compute
GRID AI
Operates Node
From GRID AI
- Funds infrastructure
- Operates and maintains infrastructure
- Compensates Node electricity
- Creates host participation value
From the Node
AI compute
to the network
Recovered heat
to the building
Compute has value.
Heat has value.
Location has value.
GRID AI connects all three.
ConceptualIllustrative architecture of the intended model. No prices, payments, percentages or returns are implied.
16 — First node
The network starts with one node.
Before a million nodes, there is one.
GRID AI is developing its first physical prototype Node to test the architecture in the real world. Its final engineering architecture is not yet fixed.
12Next
The next datacenter might already have an address.
GRID AI is building the infrastructure for distributed intelligence.