Key Takeaways
- Submer Group has spun‑out Corenix, a factory‑built, modular data‑center provider targeting AI‑heavy workloads.
- Each Corenix module follows an NVIDIA reference architecture and is fully tested (IT, power, cooling, networking) before shipping.
- The end‑to‑end delivery model covers factory integration, site‑level installation, acceptance testing and commissioning under a single contract.
- Submer’s legacy in liquid‑cooling and high‑density thermal engineering enables Corenix to support >1.5 kW per rack and hundreds of MW of aggregate capacity in under‑nine‑month lead times.
- Compared with conventional build‑outs, Corenix promises 30‑50 % faster deployment, up to 40 % lower OPEX, and a repeatable, scalable path for successive AI‑generation upgrades.
Introduction
Submer Group, a veteran in liquid‑cooling and high‑density infrastructure, announced the formation of Corenix, a modular data‑center company aimed at “neoclouds,” hyperscalers, and AI operators. The venture is led by Martin Renkis, appointed chief executive officer, who brings more than two decades of experience in data‑center engineering and global project delivery.
Modular Design‑to‑Deploy Workflow
Factory‑Level Integration
Corenix designs, builds, and integrates complete data‑center pods—including compute, power distribution, liquid‑cooling loops, networking fabric, and monitoring systems—inside a controlled production facility. Every pod undergoes functional verification (IT, power, cooling, and networking) and site‑acceptance testing before it leaves the factory.
On‑Site Commissioning
After shipment, Corenix crews handle mechanical installation, connection to the client’s utility and network backbone, final acceptance testing, and hand‑over of an operationally ready system. The entire scope—engineering, supply‑chain logistics, manufacturing, and field deployment—is managed under a single contract, eliminating the need for multiple vendors.
Technical Foundations
| Feature | Corenix Specification | Traditional Build (Typical) |
|---|---|---|
| Reference Architecture | NVIDIA DGX‑H100 reference design (GPU‑centric) | Vendor‑agnostic, mixed CPU/GPU |
| Power Density | Up to 1.5 kW per rack (liquid‑cooled) | 0.8‑1.0 kW per rack (air‑cooled) |
| Cooling Method | Direct‑to‑chip liquid cooling, 10‑15 °C ΔT | Air‑side economizers, 20‑30 °C ΔT |
| Deployment Lead Time | ≤ 9 months from order to live | 12‑24 months |
| OPEX (Cooling) | 30‑40 % lower energy cost per compute unit | Baseline |
| Scalability | Plug‑and‑play modules, incremental capacity | Site‑wide redesign often required |
The table highlights how Corenix’s NVIDIA‑based, liquid‑cooled modules deliver higher compute density while reducing power‑to‑cooling conversion losses.
Submer Group Heritage
Submer’s portfolio includes a multi‑campus deployment delivering over 300 MW of liquid‑cooled capacity in just nine months—a benchmark that validates Corenix’s promise of rapid, high‑density rollout. The company’s decade‑long expertise in thermal engineering and global infrastructure logistics underpins the new venture’s ability to meet the aggressive serviceability and upgrade cycles demanded by modern AI workloads.
Market Position & Use Cases
- Hyperscalers: Accelerated provisioning of AI‑focused pods to meet spikes in training demand.
- Neocloud Providers: Rapid geographic expansion without the capital‑intensive build‑out of traditional colocation.
- Enterprise AI Operators: Turnkey, on‑premise AI clusters that can be upgraded by swapping modular pods rather than overhauling entire facilities.
Bottom Line
Corenix leverages Submer Group’s proven liquid‑cooling and modular engineering pedigree to deliver a factory‑tested, plug‑and‑play data‑center solution built around NVIDIA’s AI‑optimized reference designs. By consolidating design, manufacturing, and on‑site commissioning under one umbrella, Corenix shortens deployment cycles, cuts operational costs, and offers a repeatable path for scaling AI compute across generations. For organizations that can’t afford the latency of conventional data‑center construction, Corenix presents a compelling, future‑proof alternative.