PrivateAIcloud
A private deployment is not a hosting choice. It is a decision about who holds the keys, who can administer the system, and what happens to your workloads if the relationship ends.
Whatprivatecanmean
The word covers four materially different arrangements. Being specific about which one you need changes the cost, the timeline and the responsibilities.
Dedicated hosted
NeuroCluster-operated infrastructure, dedicated to you. Isolation from other tenants at the compute, storage and network layer, with residency fixed by contract.
Your cloud tenancy
Deployed into your own cloud account. Your existing agreements, network controls and cost management apply; data never enters infrastructure we operate.
On-premises
Your datacenter and your hardware. Used where residency or latency requirements are absolute, or where an existing investment should be used.
Disconnected
Air-gapped, with no outbound dependency for inference, updates or telemetry. Updates arrive through a controlled process rather than a network path.
Whatyougetineveryprivateposture
- Model independence
Open-weight and private models hosted inside the boundary, with no requirement to call an external provider.
- Workload isolation
Separation between tenants, environments and agents at the compute and network layer.
- Key custody
Customer-held encryption keys, including the option to hold the root of trust yourself.
- Operator access control
Named, time-bounded and logged administrative access, subject to your approval where you require it.
- Sovereign operations
Documented runbooks so your own staff can operate, upgrade and recover the system.
- Portability
Ontology, policies, evidence and agent definitions exportable in documented formats.
- Standard Kubernetes
Conformant Kubernetes rather than a proprietary control plane, so the deployment is not a lock-in mechanism.
- Residency by construction
Placement constraints enforced by the platform, not merely promised in a policy document.
- Evidence stays local
Decision records and audit trails remain inside your boundary.
The honest trade-off
A disconnected deployment removes an entire class of risk and adds an operational burden: you own patching, capacity and recovery. That is the right trade for some environments and the wrong one for many. It should be a decision, not a default.
Whodoeswhat
| Responsibility | Dedicated hosted | Your cloud | On-premises / disconnected |
|---|---|---|---|
| Hardware and capacity | NeuroCluster | Your cloud provider | Customer |
| Platform operation | NeuroCluster | Shared | Customer, with support |
| Upgrades | NeuroCluster | Scheduled with customer | Customer-initiated |
| Key custody | Customer option | Customer | Customer |
| Monitoring | NeuroCluster | Shared | Customer |
| Incident response | NeuroCluster, with customer | Shared | Customer, with support |
Questions
- Can we start shared and move to private later?
- Yes, and it is a common path. The architecture is identical, so migration is a data and configuration exercise rather than a re-implementation. Plan it deliberately rather than discovering the need mid-programme.
- Do models work as well without a frontier API?
- For a great deal of enterprise work, yes — particularly extraction, classification, retrieval and structured reasoning over a well-modelled ontology. For some tasks a frontier model is meaningfully better. The architecture lets you make that call per class of work rather than once for the whole system.
- What is the minimum viable footprint?
- It depends on the workload mix, model sizes and availability requirements. This is a sizing conversation with real numbers rather than a published figure, and we would rather scope it properly than publish one that misleads.
- Available
- In production use today.
- In development
- Being built now. Not yet generally deployable.
- Planned
- On the roadmap. Not yet in development.
- Research
- Under investigation in Labs. May not become a product.
- Strategic direction
- Intent, not a delivery commitment.
Continue
- Infrastructure overviewThe full stack and every deployment posture.
- SovereigntyThe six control dimensions a private deployment addresses.
- ModelsRunning open-weight and private models inside the boundary.
- RuntimeThe execution layer being deployed.
- ArchitectureReference architecture for security review.
- GovernmentWhere disconnected deployment is most often required.
Bring us one operational problem.
You do not need a finished brief. Bring the problem — we will work out the next step together.
Or book a call with the team