Sovereign · Off-grid · Deployable

AI infrastructure that does not wait for the grid.

HEARTHCube brings its own power plant. Designed for islanded operation, it removes utility grid interconnection from the deployment critical path, keeps compute behind the customer’s fence line, and puts waste heat to work instead of paying to reject it.

No grid interconnection required by design. Site, fuel, safety, environmental and other applicable permitting still apply.

Assess my AI infrastructure

What must stay sovereign, what compute you actually need, and what to deploy. A paid readiness and deployment assessment.

Readiness Assessment →

Evaluate a site or heat-reuse opportunity

Site, power, fuel, thermal offtake and preliminary economics, ending in a clear go, conditional go or no-go.

Feasibility Study →

Two or three organisations

Become an ARKADIN design partner

Early access and roadmap influence on on-premises AI that shows its working and abstains rather than guesses.

Design Partner Program →

HEARTHCube · Sustainable Power & Data

The HEARTHCube difference

Power is part of the product.

Most data-centre projects begin by asking the grid for permission, capacity and upgrades. HEARTHCube begins with its own generation. For an islanded deployment, the utility interconnection queue is not the schedule.

Speed to power

No utility load interconnection is required to operate the machine as designed. That removes a major external dependency from deployment timing.

Control at the fence line

Generation, compute and thermal management travel together. The customer controls the site, the data boundary and the operating environment.

Fewer grid-specific gates

Islanded operation can remove utility interconnection studies, network-upgrade dependencies and grid-service approval from the critical path. It does not remove site, fuel, safety, emissions, zoning or other applicable rules.

What we build

Three systems, one thesis.

The infrastructure that powers AI and the infrastructure that keeps it accountable are the same problem at different altitudes. We build at both.

The vessel

HEARTHCube

A sovereign, off-grid AI data centre in two forty-foot ISO containers. Configured for roughly 131 petaFLOPs of sparse FP8 compute, with a self-contained power plant and a five-stage thermal cascade modelled to recover more than 95 percent of waste heat. It is engineered to operate without a utility grid connection.

The mind

ARKADIN

A neuro-symbolic AI architecture built on a generate-and-verify loop. It routes proposals through explicit checks, records provenance, and abstains rather than fabricates when confidence is insufficient. Designed for on-premises, air-gapped deployment and operator audit.

The silicon path

CTT-1

The Civic Tensor Tile, our pathway toward a sovereign compute substrate, integrating compute architecture with the thermal and infrastructure requirements of HEARTHCube. Its nearest-term contribution is thermal.

Six patent filings support the platform.

Why now

Three constraints, one root cause.

The grid is the bottleneck. Large-load power availability, network upgrades and interconnection processes can determine when new compute is allowed to come online. HEARTHCube brings its own generation so the utility grid is not a prerequisite to operation.
Sovereignty is becoming operationally critical. Regulated healthcare, defence and federal workloads can face security, procurement and data-governance constraints that make shared multi-tenant infrastructure unsuitable.
Heat is a stranded asset. Nearly all the energy entering a conventional facility leaves as warmth nobody captures.

The off-grid advantage

Conventional facilities begin with a utility service assumption: grid power in, waste heat out. HEARTHCube changes the starting point. Onsite generation and islanded operation remove grid interconnection from the machine’s operating dependency, while the thermal cascade turns rejected heat into useful output.

>95%

Modelled waste-heat recovery

72 hrs

Target site-to-operation

HEARTHCube improves the useful work drawn from each unit of fuel. It does not circumvent the second law of thermodynamics, and we say so before anyone asks.

How we work

Paid steps toward deployment.

Every engagement is structured as a paid step toward deployment: fixed scope, a concrete decision or deliverable, and work that moves a real infrastructure decision forward.

01

Readiness & Deployment Assessment

Workload and data classification, sovereignty constraints, cloud versus on-premises analysis, and a staged infrastructure roadmap.

Regulated enterprise
Healthcare · Defence
Legal · Finance

02

Site & Thermal Feasibility Study

Site, electrical, fuel, compute and thermal offtake, with preliminary economics. Ends in a clear go, conditional go or no-go.

On-premises AI buyers
Distributed infrastructure

03

Heat-Reuse & Thermal Integration Study

Waste-heat characterisation, recovery pathways, offtake options and the economics of putting them to work.

Data centres · Campuses
Industrial operators

04

ARKADIN Founding Design Partner

Early access, a defined use case, validation and genuine roadmap influence. Two or three organisations only.

Regulated organisations

See the engagements Start a conversation

Partners

Built with specialists.

d.i.t. · Erika Aversa · Netherlands

Hyperscale data-centre heat-exchange expertise, brought into HEARTHCube's thermal engineering and our European development strategy.

Independent review

HEARTHCube's architecture has been reviewed by an independent, Sweden-based heat-recovery specialist. We publish what has been verified and label what has not.

Honest labels throughout: targets are labelled as targets, conversations as conversations, and a patent claim filed is not a claim proven.

Founder

One founder. Three systems. One operating discipline.

Jack Stallion is the founder and engineer behind HEARTHCube, ARKADIN and CTT-1. His work spans systems architecture, technical research, commercialization, investor communication, partnerships and multidisciplinary project execution.

Before JSE, he spent nearly two decades in film and video production, executive producing two feature films. Roughrider screened at SXSW and won Best Picture at the Bare Bones Film Festival.

Hold the vision steady.

A feature film and a deep-tech platform demand the same core discipline: coordinate specialists, budgets and moving parts under hard constraint, then make the whole legible enough for others to build, fund and trust.

Investors

One machine proves the model.

HEARTHCube is in active investor discussions. Unit One is designed to replace modelled performance with measured operating data. Technical architecture, unit economics and market materials are available for qualified discussions, subject to appropriate confidentiality.

Start an investor conversation

'You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete.' R. Buckminster Fuller

Nothing on this site is an offer or solicitation to buy or sell any security.