Is cloud still actually worth it? Modern on-prem compute and local networks are practically bulletproof now.
When the cloud push started a decade ago, the selling point made complete sense: hardware was a pain to manage, SANs were expensive, disk drives died constantly, and scaling required months of lead time. Cloud promised high availability and low operational overhead.
Fast forward to today, and the reality on the ground feels completely different:
- Hardware is ridiculously reliable. Modern enterprise compute, hypervisors, NVMe storage, and 10G/40G local networking just work. You can drop a couple of redundant nodes in a rack, set up hyper-converged storage, and touch it maybe twice a year.
- Local network speed/latency beats the WAN every time. Running traffic locally over high-speed switches is instant, predictable, and doesn't rely on ISP stability or cloud provider region outages.
- The cost equation flipped. Cloud billing has turned into an endless maze of egress fees, API call costs, compute markups, and surprise monthly charges that balloon every time management blinks.
It makes me wonder: if local hardware, compute, and networking have gotten this fast, cheap, and stable, why are we still defaulting to cloud for standard workloads?
I get it for massive, unpredictable elastic scaling, global CDN edge delivery, or multi-region AI workloads. But for standard line-of-business applications, storage, and core infrastructure, running on compute you own feels faster, safer, and infinitely cheaper.
I'd love to hear from people managing infrastructure day-to-day:
- Is cloud still genuinely giving you value over local compute/networking, or are you just stuck there due to legacy decisions?
- Has anyone actually run the ROI comparison recently between modern local hardware vs. cloud compute costs?
- What are the honest dealbreakers holding you back from bringing workloads back on-prem?