Interactive calculators for the economics of running your own cloud platform. Compare five-year TCO against thirteen on-prem platforms, price a hyperscaler bill against your own hardware, size the unit economics of a hosting business, or weigh owning GPUs against renting them. Every input is editable, every price carries its source and date, and every result recomputes live — built by Ænix, the team behind Cozystack.
Platform TCO — Cozystack vs 13 on-prem platforms
The full model: five-year total cost of ownership against VMware (VCF / VVF / vSphere), Nutanix, OpenShift (OVE and Container Platform), Proxmox VE, OpenStack, CloudStack, OpenNebula, Harvester, Rancher and Virtuozzo. Three cost articles — software, one-time migration, personnel. Every default price carries its source, date and nature (vendor list, third-party, derived, owner estimate), and each comparison states where the other platform wins.
Per-platform breakdowns: vs VMware · vs Nutanix · vs OpenShift · vs Proxmox · vs OpenStack · vs CloudStack · vs OpenNebula · vs Harvester · vs Rancher · vs Virtuozzo
Cloud repatriation — hyperscaler bill vs your own hardware
Already on AWS, Azure or GCP. Price the same workload footprint — vCPU, RAM, block and object storage, managed Kubernetes, databases, GPUs, egress, cross-AZ — against running it on Cozystack on owned or rented hardware, with commitment and enterprise discounts on the cloud side and hardware, power, PUE, colocation and operations staffing on yours.
VMware exit — quick estimate
A four-input sanity check on a VMware/VCF renewal: annual saving, three-year net, migration payback. For the sourced five-year model with quote sensitivity, use Cozystack vs VMware above.
VMware-exit savings calculator
Illustrative. Edit the inputs. License line disappears (Apache 2.0); you pay support and/or the build.
Hosting-provider unit economics
If you sell managed cloud to your own customers, our full ISP unit-economics calculator models the monthly P&L — infrastructure footprint, service portfolio (managed Kubernetes, VMs, databases, GPU, object storage), capacity utilization, staffing, and a before/after Cozystack comparison — with multi-currency support and a one-click PDF report.
See the Public Cloud Platform for the product behind this model.
GPU / AI-inference ROI
Owning GPUs on your own platform versus renting equivalent GPU capacity in the cloud — with utilization driving the effective cost per GPU-hour.
GPU ROI: own GPUs vs. cloud rental
Compare the monthly cost of owning GPUs on Cozystack against renting equivalent GPU capacity in the cloud. Every input is editable.
Illustrative. Owned cost/month = amortized capex + hosting (can exceed rental during the amortization period). Capex payback measures operational cash recovery (rental avoided − hosting) and excludes amortization — once amortization ends you own the cards and pay only hosting. Cloud rental is paid regardless of utilization.
See the AI Platform and Sovereign AI.
How these calculators work
- Two kinds of tool: the TCO, repatriation and ISP calculators are full models — every price carries a source and a date, the assumptions are editable, and each produces a PDF report you can hand to finance. The VMware-exit and GPU blocks on this page are quick four-input estimates, useful for a first sanity check and nothing more.
- What they are: independent, editable estimators for the economics of running a cloud platform on an open foundation.
- Who they’re for: infrastructure, finance and procurement teams scoping a platform build, a VMware exit, a hosting business, or a GPU investment.
- The target platform: Cozystack, Apache 2.0 — no per-core/per-socket license. You pay for support and/or the build.
- Common pitfall: comparing license-to-license only, and ignoring migration cost, staffing, and the workloads that should stay put.
Turn the numbers into a plan
A discovery call turns these estimates into an honest, workload-level TCO — including what to keep where it is.
Ænix is the team behind Cozystack — a CNCF project (Sandbox today; Incubating expected late summer 2026), Apache 2.0. Ænix commercializes it as Ænix Platform as three platforms on one engine — Public Cloud, Private Cloud and AI — that combine rather than exclude each other.