400 reasons to not use Microsoft Azure

Original Article ↗ Hacker News Discussion ↗

Overall sentiment and reliability

  • Many commenters report Azure as unreliable and fragile compared to AWS/GCP: random VM shutdowns for “maintenance,” flaky networking, AKS pods losing connectivity, and managed databases with poor latency or unexpected failures.
  • Others say it “worked fine” for standard needs (VMs, storage, basic managed services) and view Azure as roughly on par with AWS when used conservatively.
  • Several long-time users describe Azure starting out okay, then accumulating breaking backend changes and regressions over years until they eventually migrated away.

UX, portal, and documentation

  • Deep split on the portal:
    • Fans praise the “single pane of glass,” hierarchical resource groups, easier global view of what’s running and costing money, and better top-down organization than AWS.
    • Critics call it slow, cluttered, hard to navigate, with tiny fonts, lots of horizontal scrolling, and basic actions taking many seconds or minutes. Some note bizarre limitations (can’t open many links in new tabs, missing creation metadata).
  • Similar split on docs: some say Microsoft has excellent, well-funded documentation; others find Azure docs incomplete, outdated, and poorly organized.

Infrastructure as Code and tooling

  • Strong frustration around Terraform on Azure: perceived as a “third-class citizen,” missing parameters, breaking changes, and much more boilerplate than AWS/GCP.
  • Azure-specific tools (ARM, Bicep) are seen by some as lock‑in and by others as necessary to get a coherent experience.
  • General advice from several: stick to “standard” primitives (VMs, containers, Postgres, object storage) and avoid proprietary Azure services to reduce pain and ease migration.

Managed services, networking, and performance

  • Multiple horror stories: Cosmos DB early versions, Azure Functions/Elastic Jobs, managed Postgres/SQL with high latency, quotas and NAT Gateway throttling, and AKS instability.
  • Networking in Azure is called out as particularly problematic: port exhaustion, cross-subscription oddities, long VPN gateway operations, and control-plane APIs that are slow or unreliable.
  • Some note that low-level core services (VMs, basic storage/queues) tend to be much more stable than higher-level “platform” offerings.

Security, support, and outages

  • Links and anecdotes highlight serious Azure security vulnerabilities over time; one commenter quips Azure excels more at security reports than security.
  • Azure support is widely described as poor: misreading tickets, slow or no fixes, and “workarounds instead of bugs being fixed.”
  • Compared with AWS and especially GCP SRE culture, Azure is portrayed as less transparent and less rigorous in postmortems.

Pricing, quotas, and billing surprises

  • Azure is often perceived as more expensive than alternatives (including AWS) for comparable compute and managed databases.
  • Some organizations choose Azure primarily because of large discounts, enterprise agreements, or co-sell programs, not because it’s technically superior.
  • Several examples of unpleasant surprises: Sentinel ingest costs triggered by chatty control-plane logs, mysterious services appearing on bills, and regional SKU unavailability forcing architecture changes.

Vendor lock‑in, ecosystem, and business drivers

  • Many argue Azure deliberately “does things differently” to lock customers in and keep them inside the Microsoft stack (Office 365, Entra, Azure, DevOps).
  • For B2B, multiple commenters say Azure can be the rational business choice: customers already standardize on Microsoft, co-selling incentives are strong, and spend can be bundled into existing Microsoft contracts.
  • There’s criticism of tactics like tying customer Office 365 discounts to vendors hosting on Azure, and of Azure’s lack of S3 API compatibility.

Azure DevOps, M365, and broader Microsoft UX

  • Azure DevOps: mixed but generally lukewarm; pipelines seen as buggy and half-migrated to YAML, boards and wiki weaker than Jira/Confluence, but some prefer its integration versus juggling multiple tools.
  • M365 admin, Intune, Entra, and other dashboards are repeatedly cited as chaotic and poorly designed; some see this as emblematic of Microsoft’s broader UX issues and constant renames.

Alternatives and self‑hosting

  • Several comments advocate for simpler setups: VPS/dedicated servers (Hetzner, DigitalOcean, Linode), or Cloudflare for small apps, arguing most workloads don’t need big-cloud complexity.
  • Others counter that cloud still wins for global reach, elasticity, compliance, and avoiding on‑prem operational burden—especially beyond a single-server scale.