AI that runs inside your own walls.
Solace AGI Sovereign AI puts the models, your data and the agents on hardware your company owns, inside your own network. Nothing leaves unless your company allows it.
Delivered and supported in Vietnam with DTS Network (DTS).
Sovereign means three things stay with you.
Most AI is rented: your questions and your documents travel to someone else's computers and come back as answers. Sovereign AI turns that around. The AI comes to your building and works there.
The models
The language models are open-weight models that run on the appliance's own processors. Your requests are answered there, not by a public AI service.
Your data
Documents, records and everything the AI reads or writes stay on servers in your own building or data center.
The agents
The agents that research, draft and file work run in the same place. A person you name approves what goes out, and evidence is logged for specific approved actions in supported workflows.
Nothing leaves unless your company allows it.
What your company gets.
Deploy AI entirely within an environment your enterprise controls, meeting requirements for security, performance and scalability.
Data stays within your infrastructure.
The deployment is built to keep personal data inside your infrastructure, to support your company's obligations under Vietnam's Personal Data Protection Law (PDPL 2025).
Costs you can predict.
The models run on hardware you own, so there are no rising API fees and the budget stays in your control.
Built for enterprise AI in manufacturing.
Each size is matched to the work: from one lab's experiments to many workflows running at once across a factory.
From one team to the whole factory.
Start where the need is clearest and add capacity as more departments come on.
An isolated (air-gapped) deployment is available.
For work that must stay off the internet, the system can run with no outside connection. It is scoped and agreed in writing for each company.
Four ways to run enterprise AI.
| Solace AGI Private AI Box | Public AI APIs | Global hyperscalers | Generic GPU servers | |
|---|---|---|---|---|
| Data sovereignty (PDPL 2025) | Yes | No | No | Varies |
| Predictable costs, no rising API fees | Yes | No | Varies | Varies |
| Understanding of Vietnamese regulations | Yes | No | Varies | No |
| Flexible deployment to fit your needs | Yes | No | Varies | No |
| Isolated (air-gapped) deployment | Yes | No | No | Varies |
✓ Yes– Varies× NoAssessment from the DTS overview of Solace AGI Private AI.
Four sizes, from one lab to a whole factory.
Each size is delivered with the Solace AGI software set up for your company. The hardware is confirmed with each company after a review of what the AI will be used for.
-
One team or one lab
Dragon Lab
For evaluation, solution development and small teams.
- GPU
- 1 × NVIDIA GeForce RTX 5090
- GPU memory
- 32 GB
- System memory
- 128 GB
- Deployment
- About a few hours
-
Specialized departments
Dragon Department
Balances performance and cost for many real-world uses.
- GPU
- 1 × NVIDIA RTX PRO 6000 Server Edition
- GPU memory
- 96 GB
- System memory
- 256 GB
- Deployment
- About a few days
-
Multiple departments and use cases
Dragon Division
More processing capacity, for several workflows at once.
- GPU
- 2 × NVIDIA RTX PRO 6000 Server Edition
- GPU memory
- 192 GB
- System memory
- 512 GB
- Deployment
- About a few days
-
A whole factory or large projects
Dragon Factory
The highest performance, for complex, large-scale AI systems.
- GPU
- 8 × NVIDIA RTX PRO 6000 Server Edition
- GPU memory
- 768 GB
- System memory
- 1 TB
- Deployment
- About a few days
The same Solace AGI, on your hardware.
Solace AGI rebuilds the software a company runs on as one workplace the company owns: agents do the work, and people approve. On the Private AI Box, that workplace, its data and its models all live on the appliance.
Solace AGI's hosted service sends agreed model calls through solaceagi.com, and its standard terms make no air-gap promise. The Private AI Box is for companies that need the models and the data inside their own network; an isolated deployment is agreed separately, in writing.
- Solace BuzzThe workplace and its agent network: the modules your teams work in, with agents inside each one.
- AI agentsResearch, drafting, data entry, reports and follow-ups, with a person approving anything that goes out.
- Solace BrowserThe browser agents use to work in the web systems you already have.
- Private data platformA database for your organization, kept on the appliance. Access checks cover customer records such as CRM records, messages and files; the remaining routes are being brought under the same check, as the Trust page reports.
- Local AI modelsOpen-weight language models served from the appliance's own GPUs.
What you own: the Ownership Covenant →How access and approvals work: Trust →Watch the demo videos: telecom and healthcare →
Delivered in Vietnam with DTS.
DTS Network (DTS) works as a bridge for technology trade between Vietnam and the United States. DTS delivers and supports the appliance in Vietnam, with the Solace engineering team behind the software.
The DTS smart manufacturing roadmap
- Assess
- Connect
- Data
- Automate
- Real value
AI & Robotics
Solace AGI Sovereign AI, with robotics and AI for manufacturing automation.
Data & Marketplace
Collecting, cleaning and processing data into dashboards.
DeepTech Hub
Experimentation and production infrastructure, and connections to partners in the United States.
A pilot is scoped for each company.
A pilot starts small: one or two uses, on one appliance, with your own people approving the agents' work. Scope, hardware and timing are agreed with each company before anything is ordered.
What we ask you to bring
- One or two uses that matter most
- A technical contact on your side
- Where it will run: your data center or server room
- Access to the data those uses need
Talk to us about AI you own.
Book a call with Phuc, or write to DTS in Vietnam. Bring the work you want AI to do and the rules your data must follow.
In Vietnam, DTS answers at contact@dtsnetwork.group. Elsewhere, write to phuc@phuc.net.