Azure Local in 2026: a buyer's guide for mid-market enterprises

Azure Local is Microsoft's hyperconverged platform for running Azure services in customer facilities. It's the right product when sovereignty, latency, or regulation pull workloads out of the public cloud. It's the wrong product when you're using it to make a hyperscaler bill feel smaller — that's what private cloud is for.

The mid-market sizing trap. Most first-time Azure Local builds we see are over-sized by 30-50% because the vendor reference architectures assume Fortune 500 workloads. A four-node cluster (the typical "starter") handles roughly 200 stateful VMs, 8-12 TB of working storage, and 50-200 AKS-on-Local pods comfortably. If you're below those numbers, you're probably looking at a colo + Azure Arc setup, not a full Azure Local cluster.

Vendor selection. Dell, HPE, and Lenovo all ship validated Azure Local nodes. We've deployed all three in production. In practice: Dell's PowerEdge line has the deepest spares pipeline and longest-running firmware track record for Azure Local. HPE's nodes are mechanically nicer (cleaner cable management) but the validated-firmware update windows are shorter. Lenovo's ThinkAgile MX is the price leader. None are bad choices — the differentiation is below 5% on TCO over five years.

Networking. The single biggest factor in Azure Local cluster stability is networking. RDMA-over-Converged-Ethernet (RoCE) with PFC and DCB tuned to vendor reference is non-negotiable. We've inherited two customer clusters where networking was "close enough" — both had recurring storage timeouts that disappeared the day we re-tuned the switches.

AKS on Azure Local. The killer feature for most mid-market buyers. Lets you run production Kubernetes on validated hardware with the Azure-managed control plane. We use it for customer-facing application workloads where the alternative was managed AKS in Azure public cloud at 2-3× the cost.

What goes wrong. Two failure modes account for 80% of post-deployment incidents: (1) Microsoft Arc enrollment misalignment with the customer's existing Entra ID forest, and (2) cluster sets being defined in ways that block storage live-migration. Both are fixable but expensive to retrofit. Get them right at deployment.

The Ultiblob recommendation. If you're considering Azure Local, the right first step is a 4-hour scoping session — not a $200k purchase order. We run free assessments for organizations actively evaluating Azure Local against alternatives. The output is a written architecture proposal, a vendor-neutral hardware sizing, and a deployment plan you can take to any partner.