Everyone told us to use the cloud. We chose Intel NUCs instead.
Not because we’re contrarian — because the economics made sense. Running a full enterprise platform on 3 mini-PCs costs roughly €1,500 one-time versus €2,000+/month on any major cloud provider. For a small DevOps consultancy, that difference is the margin between profitable and burning cash.
The Cloud Tax Problem
We needed to run our entire platform: CI/CD pipelines, container registry, monitoring stack, identity management, service mesh, and an AI agent platform. On AWS, that’s easily 10+ services, each with its own pricing model, each designed to make you forget how much you’re spending.
Quick napkin math for our workload on AWS:
- 3x m6i.xlarge instances: ~$450/mo
- EKS control plane: $73/mo
- EBS storage (1.5TB): ~$90/mo
- ALB + data transfer: ~$150/mo
- RDS for Keycloak + apps: ~$200/mo
- S3 for backups + artifacts: ~$50/mo
- Total: ~$1,000/mo minimum — and that’s before any AI/ML GPU workloads
Our 3 NUCs? One-time cost. We pay for electricity and internet. That’s it.
What We Needed to Run
This isn’t a hobby cluster. We needed production-grade infrastructure:
- 84 applications across DevOps, observability, security, AI/ML, and collaboration
- GitOps-managed — every change tracked, auditable, reproducible
- SSO everywhere — Keycloak integrated with 15+ services
- Full observability — metrics, logs, traces, profiling, error tracking
- Security — runtime protection, policy enforcement, secrets management
- AI platform — LLM gateway, agent orchestration, experiment tracking
- VM workloads — KubeVirt for running traditional VMs alongside containers, no separate hypervisor needed
The question was never “can Kubernetes run on bare metal?” — it was “can we build something that rivals cloud-managed platforms, on hardware we own?”
Spoiler: yes.
The Hardware
3x Intel NUC 13 Pro:
- Intel Core i5-1350P (12 cores, 16 threads)
- 64GB DDR4 RAM each
- 500GB NVMe SSD each
- 2.5GbE networking
Total: 36 cores, 192GB RAM, 1.5TB NVMe storage. Rack-mounted in a home lab, connected to a UPS.
What’s Coming in This Series
- Part 2: Bootstrapping — Tinkerbell, Cluster API, and automated provisioning
- Part 3: The GitOps architecture — 84 apps, zero manual deployments
- Part 4: Observability at scale — LGTM stack with multi-tenant Mimir
- Part 5: Running an AI agent platform on Kubernetes
Each post will include real configs, lessons learned, and things we’d do differently.
Cloud Native Solutions builds and operates Kubernetes platforms. Talk to us if you want this for your team.