Moore's Scofflaws

Original Article ↗ Hacker News Discussion ↗

Per‑core software licensing and slowing hardware gains are pushing more companies to question the economics of public cloud versus owning their infrastructure. Commenters contrast hyperscalers’ flexible but often expensive services with private cloud setups, arguing that beyond a certain scale, well‑run on‑premises hardware can be far cheaper—while noting the operational complexity and expertise this demands. Oxide’s integrated rack‑scale systems are debated as one attempt to bring “cloud‑like” elasticity to owned hardware, raising questions about pricing transparency, vendor lock‑in, and whether such offerings can win against both commodity servers and established enterprise vendors.

Per-core licensing & vendor lock-in

  • Many see per-core licensing as outdated and arbitrary, especially with multi-core, SMT, heterogeneous cores, and GPUs.
  • Historically, per-socket made more sense; now per-core is viewed as a late-stage extraction tactic rather than cost recovery.
  • Some vendors have complex “core equivalence” tables across architectures, but pricing is heavily negotiated anyway.
  • Moves like per-core fees and “AMD taxes” are read as signals to mitigate vendor dependence, but capital markets often reward this extractive behavior.

Cloud economics vs on-prem

  • Several commenters report that, at scale and with stable workloads, self-hosting is far cheaper than public cloud, sometimes by an order of magnitude, even including staff.
  • Examples: internal clouds beating AWS on cost; colo infrastructures massively undercutting proposed moves to managed clouds.
  • Others cite many case studies where cloud adoption produced positive ROI, especially for teams that leverage cloud-native designs and have volatile or bursty workloads.
  • Startups are said to often overpay on cloud through overengineering (e.g., chaining lambdas for simple workflows).

Oxide’s model and target market

  • Oxide is framed as “deploy elastic, buy your hardware”: hyperscale-style, rack-scale systems for organizations that want cloud-like operations on owned assets.
  • Supporters highlight integrated hardware+software, better operations, and lessons from hyperscalers and prior large deployments.
  • Skeptics question whether the product is sufficiently differentiated from commodity servers + free software, note lack of GPU/ML focus, and worry about vendor lock-in to a bespoke rack.

Pricing, sales model, and buyer perception

  • Enterprise buyers expect negotiated pricing; Oxide follows a “contact sales” model.
  • Some engineers want at least a ballpark to avoid wasting time; rough figures in the thread put a rack in the mid six figures.
  • Absence of public pricing leads some to assume “if you have to ask, you can’t afford it,” though others argue that’s a consumer, not B2B, mindset.

Private clouds, elasticity, and alternatives

  • Multiple industries (telco, oil & gas) reportedly run private clouds (OpenStack/OpenShift, Azure Operator Nexus) for specialized networking/performance needs.
  • Kubernetes and similar tools are seen as a way to decouple from proprietary cloud services and keep workloads portable across public cloud and on-prem.
  • Commenters debate whether the economics and complexity of “roll your own cloud” are worth it, with Oxide positioned as a more integrated answer where it is.