Our $100M Series B
What Oxide Sells / How It Works
- Product is described as a “cloud in a rack”: a fully integrated rack-scale system (compute, storage, networking, power, firmware, control plane, hypervisor, OS) that you buy, wheel into a data center, plug into power + network, and then consume via an API/console like a VPS provider.
- Built on custom hardware (sleds, backplane, BMC replacement, 48–54V power, big shared fans) and open-source software (illumos-derived OS, Rust control plane, custom storage, etc.).
- Several commenters liken it to a modern Sun/SGI-style vertically integrated system more than to Oracle; others say architecturally it’s closer to midrange or hyperscaler designs than to classic mainframes.
Target Customers, Pricing, and Fit
- Not aimed at small shops or homelabs; current SKUs are half- and full-racks (1024–2048 cores, tens of TiB RAM), with rough estimates of $400k–$1M per rack plus annual support.
- Seen as attractive for:
- Large orgs that are “all in” on public cloud but now see cost/sovereignty issues and want on‑prem “cloud-like” operation.
- Enterprises and governments that never went to cloud and want a more coherent, API-driven stack than ad‑hoc “pizza box” fleets + VMware.
- Some worry GPUs are missing in an AI-driven market; others note many workloads are still CPU-only and expect GPU or accelerator options later.
Cloud vs On-Prem Economics and Lock-In
- Many argue public cloud is very expensive for steady, predictable workloads (especially GPUs), citing examples of big savings from repatriation.
- Others counter that people often under-account for on-prem staff, capacity planning, and failures; hyperscalers’ efficiency and automation are real.
- On lock-in:
- Pro-Oxide: guests are standard VMs, stack is open source, no software license fees; migration is “just” moving VMs.
- Skeptical view: management stack + custom hardware is effectively vendor lock-in, similar in spirit to mainframes or “iPhone of the data center”; if Oxide fails, you’re on an island.
Compensation and Hiring Model
- Oxide pays almost all employees the same salary (~$207k USD, location-agnostic); sales roles have lower base + commission; no bonuses.
- Equity is explicitly not equal: earlier employees get more; equity is used to compensate for risk and varies by timing and (implicitly) role.
- Debate:
- Supporters like the transparency, reduced negotiation, and lack of performance-review games.
- Critics see focus on base salary as distracting from total comp; worry equity distribution could be uneven and opaque, and flat salary plus commission carve-outs undercut the egalitarian narrative.
- Hiring: heavy emphasis on long-form written materials (no early “screen” call; interviews only at the end). Some found this deeply useful as career reflection; others were frustrated by multi‑month waits and terse rejections.
Engineering, Cooling, and “First Principles” Claims
- Oxide emphasizes rack-level power conversion (to ~54V), bus bars, and large 80mm fans, claiming big drops in fan power and improved density vs 1U systems; they argue earlier DC designs leave a lot of efficiency on the table.
- Some commenters note similar ideas existed in blade systems and OCP; Oxide staff respond that they ended up diverging from OCP and designed mechanics, power, and firmware from scratch.
- Debate over how much efficiency you can really gain given CPU/GPU power dominates, but many appreciate the end-to-end, vertically integrated engineering focus.
Strategy, Exit, and Culture
- Round is led by a “national-interest” growth fund; participants speculate this aligns well with defense/government demand for on-prem, high-sovereignty compute.
- Oxide explicitly talks about building a “large, durable, public company” and commenters assume IPO is the target; some worry public or PE ownership will eventually erode current compensation and culture.
- Culture elements praised: open RFD process, book-club/podcast habits, strong writing, and a Sun-like engineering ethos. A minority expresses concern that attention to politics or internal ideology (e.g., social media choices, equal-pay symbolism) could distract leadership from customer problems.