Sovereignty
You decide which models run, what data is processed, and how far it goes. Sovereignty without isolation.
The case for private AI
Meet your new private AI. It's about protecting your data and what you build.
Public cloud is useful: for quick tests, occasional spikes, for what's not sensitive. The question is what happens when AI stops being an experiment and becomes part of daily operations in your development or at an institution. That's where a private node changes the rules.
Comparison
Token-based public cloud vs. your new Xeretron private AI node.
| Criterion | Cloud API | Xeretron on your infrastructure |
|---|---|---|
| Where processing runs | Provider data centers | Your organization's servers |
| Cost model | Per token consumed | Upfront investment + operations |
| How spend behaves | Grows linearly with usage | Flattens as volume increases |
| Model choice | Provider catalog | Xeretron-owned models you select |
| Price changes | Unilateral | Under your control |
| Availability | Depends on the internet and the service | Runs on local network per architecture |
| Audit | Provider reports | Your own security and inference logs |
| Resulting asset | No infrastructure ownership | Installed, depreciable capacity |
| Effort curve | Low at the start | Requires evaluation and guided deployment |
| Best for | Tests, spikes, non-sensitive workloads | Steady use, sensitive data, local governance |
Four reasons
Why institutions choose to govern their own AI.
Predictable cost
With steady workloads, cost becomes an investment decision—not a monthly surprise.
Resilience
Local operation doesn't depend on an external service being up right now.
Dedicated compute
GPU dedicated to your organization—no sharing usage limits with anyone else.
Node economics
The break-even point, explained without hype.
With the cloud, monthly spend rises in proportion to usage. With Xeretron, the curve is upfront capital plus monthly operations (power and support). Where the two curves cross is break-even.
meses_de_equilibrio = CAPEX ÷ (costo_cloud_mensual − ops_mensual)
- Indicative SMB blended profile: ~$1.50 input and ~$6.00 output per million tokens.
- Installation investment by plan from $599 to $7,999 USD depending on scope (Personal from $599 for a limited time).
- Annual license and update maintenance from $299 to $1,399 USD depending on plan.
Disclaimer. All figures are illustrative estimates. They depend on actual usage, hardware, local power, support contracted, and models chosen. We don't publish guaranteed savings.
Official pricing by plan
| Plan | Installation | Maintenance/year |
|---|---|---|
| Personal | $599 (regular $1,299) | $299 |
| Starter | $2.999 | $799 |
| Institution | $4.999 | $999 |
| Business | $7.999 | $1.399 |
Official installation prices in USD. Each additional user above the plan limit adds $29 USD/year; additional GPU license $99/year.
Technical honesty
When Xeretron is NOT a fit.
Better to say it here than to find out on an invoice or a stalled project.
Very low or sporadic use
If you query only a few times a month, a cloud API will still be cheaper and simpler.
No technical owner at all
We train and support you, but the organization must designate someone to care for the node.
Specific closed models
If you need exclusively a vendor's proprietary model, that doesn't install on-premise.
Hybrid is valid too. Many projects keep private workloads on Xeretron and use public cloud for one-off experiments. The choice is yours, and the panel is designed to make it explicit.