Interactive estimator
DC Estimator: cloud, sovereign and hybrid AI data centres
Interactive estimator for government cloud, sovereign and hybrid AI data centres: AI compute demand, compute, network, storage, power, cooling, software, facility and resilience sizing with vertical and horizontal scaling, build-buy-share-rent, TCO, investment and PPP case, sovereignty and 7R migration.
Demo scenario on a fictional country. Public reference prices with sources and dates; discounts entered by users stay in their browser. Built by Marius Mihail Russo-Got.
Inside the demo
- Scenario & deployment model — Choose sovereign, hybrid or public-first, the planning horizon and the data classification; the placement policy…
- Advisory scope coverage — How each area of a government cloud, data-centre and AI-infrastructure advisory engagement maps to a module of the…
- Cloud readiness diagnostic — Eight dimensions scored from initial to optimised against a target: strategy, legal and classification, security and…
- Migration planner (7R) — Application-by-application 7R dispositions — retire, retain, rehost, relocate, replatform, repurchase, refactor — with…
- AI compute demand — Training, fine-tuning, inference and shared GPU pools turned into GPUs from FLOPs, model memory and utilisation, then…
- 1 · Compute — Accelerators, HBM, host CPUs, 8-GPU nodes and NVL72 racks, and virtualisation hosts, each with vertical and horizontal…
- 2 · Network & fabric — Scale-up NVLink domains, InfiniBand or Ethernet RoCEv2 fat-tree fabrics, NICs, optics, front-end and storage networks,…
- 3 · Storage — Parallel file systems on NVMe for datasets and checkpoints, object storage, vector and KV-cache stores, block and file…
- 4 · Power & backup — Grid connection, transformers, UPS and distribution, generators and battery storage for AI load transients, with N, N+1…
- 5 · Cooling & thermal — Air, rear-door heat exchangers, direct-to-chip liquid cooling and immersion chosen by rack density, with PUE, WUE, heat…
- 6 · Software & operations — Kubernetes with GPU scheduling, Slurm, collective libraries and inference runtimes, telemetry and self-healing,…
- 7 · Facility & resilience — Racks, white space, build or colocation, Tier and EN 50600 class, a second site for disaster recovery, RTO and RPO.
- Sovereignty & placement — Which capabilities need domestic control, which can be sourced regionally or globally, sovereignty controls and the…
- Build · buy · share · rent — Cost per GPU-hour against utilisation for a national AI data centre, colocation, a shared regional facility and GPUaaS,…
- Deployment architectures — Cloud, centralised national, distributed regional or edge and Small AI per use case, from latency, connectivity, data…
- Operating model, investment & PPP — GOCO, PPP or state-operated: who absorbs accelerator depreciation, who hires scarce engineers, how sovereignty is…
- Price book & discounts — Public reference prices with source and date for GPUs, networks, storage, construction, colocation, energy, cloud and…
- TCO & comparison — Annual cost by category, total cost of ownership, unit costs per GPU-hour, per million tokens and per VM, and the bill…
- Compliance & sustainability — Checks against the EED and the EU data-centre rating scheme, EN 50600, NIS2, GDPR, the AI Act and sovereign cloud…
- Report & exports — Printable summary of the scenario, scenario files and CSV exports for financial models.
- Method, factors & sources — How the estimator calculates, editable engineering factors, accelerator specifications and every source behind the…
- Advisor Q&A — Questions governments, development partners and investors ask about cloud, data centres and AI infrastructure, answered…
- Situation assessment (as-is) — Baseline of today's estate — server rooms, PUE, utilisation, age, support status, staff and licences — against the…
- Tranches, lifecycle & refresh — Three investment horizons, demand-gated GPU tranches with release gates, rolling refresh and cascade from training to…
- Government engagement playbook — Mandate and operating model, site and grid feasibility before the tender, sovereign architecture zones, outcome-based…
- Consulting & services — Advisory services from diagnostic to operations transition, with deliverables, team and effort scaled to the scenario.
- Implementation flow — Ten stages from mandate to operations with entry criteria, deliverables, gates and a scenario-based timeline.
- Risk register — Scenario-driven risks for a government AI data centre with probability, impact, mitigation and owner.
- eLearning — Short courses with quizzes on AI data centres, sizing, power and cooling, sovereignty, build-buy-share, lifecycle,…
- Technical terms — Glossary of the technical terms used in the estimator: compute, network, storage, inference, power, cooling, software,…