This site runs without analytics or fingerprinting. One choice determines whether web fonts are pulled from Google Fonts — an external request that carries your IP to a third party.
ConsentExternal web-font request only. Declining swaps in system typefaces; everything else is identical.
RetentionOne strictly-necessary cookie records this choice for 12 months — required to remember it. No analytics, no fingerprinting, nothing else.
Your rightsAccess · rectification · erasure · portability. We hold no data, so there is nothing to retrieve.
NO TRACKING · NO ANALYTICS · ONE CONSENT-RECORD COOKIE (12 MONTHS) · NOTHING ELSE
Customer-controlled AIOutside models by permission
HALO AI · SMALL-BUSINESS EDITION INCLUDED
Your hardware. Your documents. Halo AI.
Halo AI combines Foxit document processing, a local AI model and approved tools to help your business find answers and get work done.
Local AI
A model selected for your hardware and the work you want to do.
Your business knowledge
Intelligent document processing (IDP) turns approved documents into knowledge your AI can use, with links back to the sources.
Work under your control
Ask questions, prepare drafts and connect approved tools. You set the permissions and review important actions.
Foxit document technology is part of the planned Halo AI integration. Your quote confirms available capabilities, licences and delivery responsibilities.
Complete systems include the small-business software edition. Document preparation, custom workflows, integrations and ongoing support are scoped separately.
From one useful workflow to an agreed setup.
Choose the work
Choose a recurring task and define the result.
Review the value
Agree how to measure time saved, quality and review effort.
Set the boundary
Choose approved sources, tools and human approvals.
Plan your Halo AI
Confirm deployment, setup and support for the agreed job.
Own the AI your business depends on.
Keep essential work on your hardware. Use outside AI when it adds value—and only with your permission.
See Halo AI at work.
Three illustrated business examples. Representative scenarios, not customer testimonials.
A local model is one part of a sovereign setup. Validation must also cover data flows, permissions, keys, updates, outside services and continuity on the actual deployed system.
SOVEREIGNTYBOX HARDWARE · FROM CA$5,000
SovereigntyBox hardware. Ready for your AI workloads.
Start with an SBX AI Edge. Bring your team's files and Halo AI together on SBX AI NAS + GPU. Move to an SBX server when the workload needs more capacity. We confirm the exact hardware, model fit, delivery and support before you order.
SBX Workstations
SBX AI Edge A2AMD STRIX HALO · 128 GB SHARED MEMORYCompact SBX AI hardware for local inference and document work. NVIDIA GB10 alternatives are also available to plan.Built to order · October 2026From CA$5,000Provisional Canadian planning price · final quote confirms landed componentsView system →Product specifications
SBX Storage
SBX AI NAS + GPUSHARED STORAGE · NVIDIA GPUSBX AI storage and GPU hardware for shared files and local inference.Configuration requires qualificationCA$15,000–30,000Entry planning range · GPU and drives specified in quote · higher-capacity systems quoted separatelyView system →Product specifications
SBX Servers
SBX AI 2U2U RACK · GPU SERVERSBX AI server hardware for shared inference and document processing, sized and tested for your workload.Configured to orderCA$35,000–70,000Provisional Canadian planning range · final quote confirms landed componentsView system →Product specifications
Hardware planning prices include the small-business edition of Halo AI with complete systems. Your quote confirms configuration, software capabilities and delivery. Document preparation, custom integrations and ongoing support are scoped separately.
Start with one useful automation. The Halo AI helps define the work, test the economics and plan a system around approved business knowledge. Run your own personalized Halo AI on the AI hardware you control; access the other AI only where your policy permits.
Start with the workflow. Size the hardware against the real load.See AI hardware →
One person can run a growing business with an Halo AI operating system.
Use Halo AI to operate the repeatable work across supplier onboarding, compliance, product administration, e-commerce, wholesale, orders, customer service and back-office follow-up. Halo AI reads the documents, works through approved systems and brings the consequential calls back to you. Start with one profitable bottleneck. Keep adding roles until the same people can run a much larger business.
THE ENTREPRENEURIAL MULTIPLIER
One person can direct a full Halo AI operating system.
The Compliance Halo AI gets product ready. The Commerce Halo AI keeps the company store, wholesale portal and channel listings current. The Service Halo AI handles routine customer and partner traffic. The Operations Halo AI coordinates suppliers, 3PLs, orders and returns. The Finance Halo AI prepares the administrative work. You set the rules, approve the important actions and keep the relationships.
ONE BUSINESS · MANY HALO AI ROLES · ONE ENTREPRENEUR IN CONTROL
2026 · THE DEPENDENCY QUESTION IS OPERATIONAL
In 2026, outside technology decisions become operating decisions.
Trade measures, export controls, lawful-access powers, provider terms and model availability can change outside the customer’s control. SovereigntyBox keeps essential Halo AI available on customer-controlled capacity while preserving approved access to outside AI when its capability justifies the dependency.
Compare workload volume, model fit, hardware, support, power and permitted outside inference. Token cost is one input to the decision—not the business case by itself.
REFERENCE OWNED CAPACITY · PLANNING ESTIMATECA$35K / YEAR
COMPARED WITH THE SAME ILLUSTRATIVE TOKEN VOLUME · NOT A QUOTE OR ROI FORECAST
Claude Fable 5FRONTIER · $10/$50 PER MCA$579K
GPT-5.6 SolFRONTIER FLAGSHIP · $5/$30 PER MCA$331K
Claude Opus 5FRONTIER · $5/$25 PER MCA$290K
GPT-5.6 TerraFRONTIER · CUT 20% JUL 30CA$132K
Gemini 3.1 ProFRONTIER · $2/$12 PER MCA$132K
Claude Sonnet 5MID · $2/$10 INTRO TO AUG 31CA$116K
Claude Haiku 4.5SMALL · $1/$5 PER MCA$58K
SovereigntyBox on-prem70B-CLASS OPEN WEIGHTS · OWNED FLOORCA$35K
GPT-5.6 LunaNEAR-FRONTIER · CUT 80% JUL 30CA$13K
The comparison shows why utilization and model choice matter: some rented tiers cost far more than the reference owned capacity, while a low-cost API tier can cost less. On-prem figure = default build amortized 5 years including support, power and operations (planning estimate, not a quote). API rates per provider pricing pages, Jul 2026. A real decision must also account for workflow labour, integration, reliability, privacy and service requirements.
Hardware for your on-prem Halo AI. Automate your business.
Size the system against a real workflow, not an AI ambition.
We configure proven hardware around the model, volume, data boundary and operating requirements identified in the plan. A completed unit can include a documented bill of materials, provenance record and deployment evidence appropriate to the agreed build. Delivery is direct or through local operators and MSP partners.
Fit
Workflow · model · volume · operating boundary
Build
Proven bases, configured and validated to order
Evidence
Available HBOM · provenance · configuration record
Distribution
Direct · local operators · MSP partners
Boundary
Connected · gated hybrid · disconnected where supported
Configure the system →Move from workload and boundary to a capacity and cost planning document.
FIG. 03-B — UNIT, EXPLODED · SCROLL TO DISASSEMBLE
THE PRODUCT FAMILY
SovereigntyBox Halo AI.
Halo AI, built from your data — owned by you. A Commerce Halo AI should not see every compliance record. A Customer Service Halo AI should not be free to publish marketing copy or change infrastructure. Each domain Halo AI is built around an approved corpus, specific instructions, permitted tools and an escalation path.
The public Halo AI helps qualify the first automation. A customer-owned instance can become the client’s AI solutions consultant—using approved local evidence to propose, but never autonomously commission, the next domain Halo AI.
L1CUSTOMER-CONTROLLED INFRASTRUCTURE
SOVEREIGNTYBOX · CUSTOMER SYSTEM · EXPLICIT HYBRIDBOUNDARY
UNAPPROVED TOOLS00HUMAN ESCALATIONON
FIG. 04-A — HALO AI KNOWLEDGE, PERMISSION & EVIDENCE ORBIT
A COMMON STARTING POINT — INTELLIGENT DOCUMENT PROCESSING
Many Halo AI start with your documents.
Contracts, policies, records, product data and operating procedures are where much of a business already explains how work should happen. Intelligent document processing turns approved material into structured, cited knowledge that a domain Halo AI can retrieve under the customer’s access rules.
04B INTELLIGENT DOCUMENT PROCESSING
THE FOUNDATION LAYER
SOURCEIngestContracts, filings, policies and records enter source-registered.
STRUCTUREExtractText, tables, fields and references become usable structure.
GOVERNClassifyMatter, policy, record and sensitive data are typed and scoped.
KNOWLEDGEIndexAn authorized corpus is indexed inside the chosen operating boundary.
HALO AI USERetrieveThe Halo AI retrieves only from knowledge permitted for that task and user.
EVIDENCECited answersEvery answer returns to its source and can be attested.
BUILT ON TOPDOMAIN HALO AI
RETRIEVE · ADVISE · ACT · CITE
PRINCIPLE IDP TURNS APPROVED BUSINESS MATERIAL INTO THE KNOWLEDGE FOUNDATION A DOMAIN HALO AI CAN USE.
The operating boundary is explicit: sources, permissions, permitted model routes and actions are defined for the deployment. Research, review and operational workflows can share the same pattern without sharing unrestricted access. Watch the concept console →
Taiwan builds it. You own & control it.
Every serious box on earth is built through Taiwan — that risk is common-mode, shared by every vendor. Ours simply removes the foreign control layer stacked on top of it. This is an anti-dependency position, not an anti-anyone position: every origin is documented, and the buyer sets what counts as acceptable.
05A CONVENTIONAL CHAIN
REMOTE AUTHORITY REMAINS
ORIGINTaiwan fabFabrication is common to every serious system.
ASSEMBLYIntegrationThe unit is assembled and tested.
CONTROLForeign layerUpdates, telemetry and release authority remain external.
DESTINATIONYour deploymentLocally sited. Not wholly under local authority.
RESULT YOU OWN THE BOX. A FOREIGN PARTY CAN STILL AFFECT THE SYSTEM.
05B SOVEREIGNTYBOX CHAIN
OPERATING AUTHORITY IS LOCAL
ORIGINTaiwan fabPer-part origin and jurisdiction are documented.
EVIDENCEIntegrate + attestThe unit carries a signed HBOM and provenance record.
AUTHORITYLocal operatorKeys, support and update authority stay in jurisdiction.
DESTINATIONYour deploymentCustomer-owned, disconnectable and locally governed.
RESULT YOU OWN THE BOX AND THE AUTHORITY REQUIRED TO OPERATE IT.
See what is running. Prove who controls it.
The Sovereignty Report gives you the answer first: external reach, local authority, configuration drift and any dependency that still needs a decision. The signed evidence remains underneath for procurement, security and assessors. This specimen shows the format; every deployed system produces its own record.
§ — NEWS
NEWSCURATED
Curated external reportingOpen any story inside its regional edition, with related reporting arranged around it.
00:00:00 UTC
FEATURED SBX WORKSTATION
SBX AI Edge A2
AMD Strix Halo 128 GB shared memory
From CA$5,000
Planning price · final configuration quoted.
View system →
Put one useful automation into production.
Start with a workflow whose value can be observed. Ground a Halo AI in approved knowledge. Set the boundary and permissions. Then configure only the infrastructure the work requires. Sovereignty is not a separate feature added later; it is how the automation is assembled and operated.
SBX AI Edge A2
Compact SBX AI hardware for local inference and document work. NVIDIA GB10 alternatives are also available to plan.
Built to order · October 2026From CA$5,000Provisional Canadian planning price · final quote confirms landed components
View system →
Product specifications
SBX AI NAS + GPU
SBX AI storage and GPU hardware for shared files and local inference.
Configuration requires qualificationCA$15,000–30,000Entry planning range · GPU and drives specified in quote · higher-capacity systems quoted separately
View system →
Product specifications
SBX AI 2U
SBX AI server hardware for shared inference and document processing, sized and tested for your workload.
Configured to orderCA$35,000–70,000Provisional Canadian planning range · final quote confirms landed components
View system →
Product specifications
