Cloud Exit Calculator

How much could you save by leaving the cloud?

Enter your current cloud spend and workload details to estimate your monthly savings running on colocation or on-premises hardware instead of AWS, GCP, or Azure. Cost is only part of the story — co-locating services on the same machine or cluster also eliminates inter-service network hops, dropping latency by 10–100× for internal calls.

Your current cloud setup

$

Combined spend across all cloud providers

Engineers managing infrastructure

Data transferred out per month

Workload types

Estimated savings

Enter your monthly cloud bill to see your estimate.

Latency benefit

→ Same machine: loopback calls drop to ~0.05 ms — vs 1–10 ms cross-AZ in the cloud.
→ Same cluster: intra-cluster traffic stays at ~0.1–0.5 ms with no VPC overhead, NAT, or load balancer in the path.
→ Database queries: co-locating app and DB on the same host or LAN cuts round-trip time from 2–8 ms to under 0.2 ms — a compounding gain for query-heavy workloads.
→ Microservices with 10+ hops: per-request savings of 50–200 ms are common, directly improving p99 response times.

Thinking of leaving the cloud?

The hardest part of cloud exit isn't the hardware — it's deployment. Kamal makes it as simple as pushing to a managed platform, without the managed platform bill.

Kamal open source

Deploy web apps anywhere — kamal-deploy.org

Kamal is "Capistrano for containers." It handles zero-downtime deploys, rolling restarts, and accessory service management across any servers you own — bare metal, colo, or cloud VMs — with no Kubernetes complexity and no vendor lock-in.

Zero-downtime deploysRolling restartsAny serverDocker-basedNo lock-in
→
Methodology & assumptions

Hardware amortization (20% of cloud bill/month): Assumes on-prem hardware costs roughly 5× less than equivalent cloud compute over a 36-month lifecycle — a conservative estimate consistent with published cloud-vs-colo studies.

Colocation & power ($300–$900/month): Base rack space, power, and cross-connects at a typical US colo facility, scaled by number of workload types.

Staff overhead ($50/engineer/month): Marginal ops time for bare-metal management. Full-time ops staff are assumed to already exist in both scenarios.

Bandwidth savings ($0.085/GB): Cloud egress costs ~$0.09/GB on average; colo bandwidth runs ~$0.005/GB on a committed port. The delta is credited against on-prem costs.

This is a rough estimate, not a financial projection. Actual savings depend on negotiated hardware pricing, colo contracts, staffing, and workload characteristics. Use this as a starting point for a deeper analysis.