Capitol AI

Sovereign Edge

Private AI. Ready to run.

The Capitol platform, dedicated hardware, and local models selected for your workload. Supplied and installed by Capitol. You own the hardware; we maintain the system.

Exploded isometric view of Capitol Edge, with the platform, local models and compute hardware floating above a protective appliance case.
Exploded isometric view of Capitol Edge, with the platform, local models and compute hardware floating above a protective appliance case.

The Capitol platform, fully local.

Search local collections, answer questions against your records, and turn findings into briefs and reports. Save the process as a workflow your team can run again.

Keep sensitive work inside.

Prompts, source documents, and outputs stay within your environment. The system sends no telemetry outside your boundary.

Run without a connection.

The platform and models run locally, without cloud APIs or internet access. Air-gapped operation is part of the system.

Models selected for your workload.

Capitol curates open-weight models for your tasks and hardware, reviews their licenses, and checks capabilities during installation and updates.

From installation to ongoing operation.

Included at deployment

Site assessment, dedicated hardware, platform and model setup, acceptance testing, and operator training. We size the system around your users, workload, power, and space.

Annual managed service

The Capitol platform license, software and model updates, scheduled maintenance, hardware warranty management, and local capacity reporting.

Updates stay offline.

Signed software and model updates arrive on encrypted media. Each release is tested before use, with rollback available if needed.

One platform. Three sizes.

From portable cases to facility racks, we match the hardware and models to your workload.

Capitol AIIllustrative form factor only. This drawing does not specify the final enclosure, equipment layout, or capacity.
Transit case

A compact system for smaller teams and disconnected field sites.

Capitol AIIllustrative form factor only. This drawing does not specify the final enclosure, equipment layout, or capacity.
Site rack

A shared system for multiple teams working across local collections and workflows.

Capitol AIIllustrative form factor only. This drawing does not specify the final enclosure, equipment layout, or capacity.
Facility system

A facility deployment for broader access and larger workloads.

Configuration and performance are validated against your workload during scoping and acceptance testing.

See when ownership costs less.

Compare your model API spend with hardware, managed service, and estimated electricity over time.

For the workload you would move to Edge.

Compare over
Edit assumptions

Power uses 70% of rated draw and 730 hours per month. Costs remain constant. Changing system size restores its hardware, service, and power assumptions.

You could save

$51,236

over 3 years compared with model APIs.

Upfront cost recovered in month 31.

Total cost over 3 years

Model APIs$729,000
Sovereign Edge$677,764
Total cost over time
Capitol AIModel API cost rises from zero to $729,000. Edge starts with $299,000 in hardware and installation and reaches $677,764 including managed service and electricity. Payback occurs in month 31.$0$375k$750kToday1 yr2 yr3 yr
Sovereign EdgeModel API

Planning estimate, not a quote. Includes hardware, service, and estimated electricity. Excludes site work, staffing, tax, financing, and discounts. Workload, output quality, and capacity require validation.

Deployment questions

Does it need internet access?

The platform and models are designed to run locally in an air-gapped environment. Signed software and model updates are delivered on encrypted media.

What do we own, and what does the service cover?

You own the hardware. The annual service includes the Capitol platform license, updates, scheduled maintenance, hardware warranty management, and local capacity reporting.

Which models can we use?

Capitol configures a curated set of open-weight models for your workload. Availability depends on hardware capacity and license review.

How do we choose the capacity?

We assess the workload, users, power, and site requirements. The configuration is confirmed during a site survey and tested during acceptance.