Oracle Halved Its Free ARM Cloud. Here's the $0-vs-$5 Kubernetes Math Now.
Oracle cut its Always Free Ampere allowance from 4 OCPUs/24GB to 2/12 in June 2026 and started terminating over-limit instances in August. A worked comparison of the most always-on Kubernetes still available for $0 versus a flat ~$5 Hetzner box, with the reclaim, capacity, and grandfathering catches priced in.
Hetzner's €5.99 Server Is a Number You Can't Buy: What the Missing Cheap Tier Does to the Self-Hosting Price Anchor
Hetzner's €5.99 cloud server is still on the price list but every Cost-Optimized model is marked unavailable, forcing new deployments onto tiers costing 3.34× to 4.24× more. This post traces the two 2026 price hikes, the DRAM shortage behind them, and how to re-anchor self-hosting budgets to the catalog you can actually order from.
Your Kubernetes Fleet Runs at 8% CPU — Here's What That Costs on Metered Cloud vs. Owned Hardware
Cast AI's 2026 report puts average Kubernetes CPU utilization at 8%, down from 10%. A line-by-line comparison of what that waste costs on metered cloud versus flat-rate dedicated hardware — with the sensitivity analysis that keeps the math honest.
Hetzner Changed Two Numbers Your Fleet Depends On: LB Idle Timeout and Assigned-IP Deletion
Hetzner Cloud made LB HTTP idle timeouts tunable (30–300s, up from a fixed 50s) and banned deleting assigned IPs. What both changes mean for ingress tuning and IP-failover runbooks on a self-hosted fleet.
CloudNativePG Restored 4.5TB of Postgres in Two Minutes: What Snapshot Recovery Really Buys Your Self-Hosted DR Promise
CloudNativePG restored a 4.5TB Postgres database from a volume snapshot in about two minutes. Here is what that RTO number means for teams running tenant Postgres on their own Kubernetes fleets, and where object-store backups still carry the region-loss half of the promise.
When Will EKS, AKS, and GKE Actually Support Kubernetes 1.37?
Kubernetes 1.37 shipped August 26, but EKS still doesn't offer it, AKS won't GA it until October, and only GKE has a build. A date-by-date scoreboard of managed version lag — and what the wait costs teams that want scale-to-zero, rootless kubelets, and native histograms now.
Tau Skipped Kubernetes Entirely: What a Wasm-Native PaaS Gains — and Gives Up
Taubyte's Tau runs a self-hosted PaaS on WebAssembly and libp2p with zero Kubernetes. This comparison tables what deleting the control plane buys — sub-millisecond cold starts, no etcd — against what it forecloses: no Docker images, no operators, no GPU scheduling.
Qovery's September 2026 Price Table Says Render $7, Railway $5, Fly.io $2.02 — Here's the Second-Year Bill It Hides
Qovery's September 2026 comparison lists Render from $7, Railway from $5, and Fly.io from $2.02 — but a typical production app bills 9–17x those headlines by month twelve. A line-by-line second-year ledger for all three, plus the flat Hetzner box.
RunPod's Scale-to-Zero vs Lambda's Pay-for-the-Rental: The GPU Billing Math That Decides Whether to Rent or Own Inference
RunPod bills GPU inference per active second with zero idle cost, while Lambda Labs rents the whole instance by the hour. A duty-cycle cost model built on 2026 rates shows exactly where each wins, and where owning a Hetzner GPU box beats both.