Service · Architecture
P7Private Deployment Architecture & Sizing.
Private AI works when the architecture matches the workload. In three to five weeks we model your workload, specify the hardware and topology, run TCO against continued API spend, and hand you a procurement-ready specification that survives the CFO conversation.
Outcomes
What changes when the engagement lands.
Right-sized architecture
Hardware and topology chosen against measured workload, not vendor pitches.
TCO the CFO can defend
Real cost of ownership against continued API spend. Decision framed in numbers.
Topology aligned to risk
Cloud, on-premise, or air-gapped chosen against sovereignty and latency requirements — not sales pressure.
Procurement-ready
A specification your procurement team can go to market with today.
Deliverables
What's in the engagement.
A three- to five-week architecture engagement that models your workload, specifies the GPU and hardware, chooses topology across cloud, on-premise, or air-gapped, models the total cost against API spend, and hands over a procurement-ready specification.
Workload model
Traffic patterns, latency budget, concurrency, and growth curve modelled and documented.
Hardware specification
GPU class, memory, network, and storage specified against workload. Ranked options with rationale.
Topology recommendation
Cloud vs on-premise vs air-gapped, chosen against your specific risk and sovereignty requirements.
TCO model
Three-year total cost of ownership vs continued API-based spend. Sensitivity analysis included.
How we deliver
Fixed scope. Named phases. Duration on the cover.
The engagement is priced against the outcome, not open-ended hours. Every phase has a duration, a named deliverable, and a check-out.
Total duration3 to 5 weeks
Sales cycle4 to 8 weeks
- 011 week
Workload discovery
Current AI usage instrumented. Growth projection agreed. Latency and availability requirements defined.
- 021 to 2 weeks
Architecture design
Hardware options modelled. Topology assessed. Reference architectures produced.
- 031 week
TCO modelling
Cost of build vs API spend modelled with sensitivity analysis.
- 041 week
Read-out and procurement handover
Specification handed to procurement. Executive read-out. Vendor conversation support if needed.
Built for
Buyers this engagement fits.
Typical buyer
CTO, CIO, Head of Infrastructure
CTOs weighing 'move to private AI'
The board wants to know the number. This produces the number.
CFOs asked to approve a large infrastructure spend
The engagement gives you the TCO analysis to approve or reject with confidence.
Regulated businesses with sovereignty constraints
You know you need private. This tells you what private looks like in your case.
Related services
Where this leads next.
Private LLM Deployment
Inference stack on your infrastructure. Optimisation. Identity, network, observability. Runbooks.
Air-Gapped Deployment
Everything in P8, plus offline install and update. Dependency mirror. Isolation validation. DR.
Sovereign AI Readiness & Governance Baseline
Governance review. Model and data inventory. Control set mapped to ISO 42001. Certification roadmap.
Talk to us.
45 minutes on your operation and the engagement you have in mind. No pitch, no deck.
Lead intake
Request a briefing
A 45-minute call. No pitch, no deck. We ask the questions we'd ask a Discovery client and tell you honestly whether this is the right next move.