The Custom Cloud AI provider.
OXMIQ provides neo-cloud services to build and run AI factories, end to end.
OXMIQ has two customers that are building multiple AI factories across the US, India, Thailand, Malaysia and Australia.
Custom cloud services.
From day zero to operation.
-
01
Feasibility
Determine the site and cluster: power, space, cooling, fabric and storage.
-
02
Architecture
Topology and intent: power train, cooling, white space, pod definition and scale units.
-
03
Design
Buildable detail: rack elevations, port maps, cable schedules and BOM freeze.
-
04
Procurement
Specification to purchase order: bid levelling, lead times, logistics and duty.
-
05
Deploy
Rack and stack, commissioning, fabric turn-up, benchmarking and acceptance.
-
06
Operation and management
Performance and thermal trending, firmware lifecycle and escalation ownership.
The operating system for AI factories and agents.
OxFactory runs every layer of the AI factory, from bare-metal tools to the user interface.
- User interfaceMission Control · Capsule CLI · Web · Native · Terminal · NVIDIA AI Workbench
- Agent frameworkNemotron Claw · ComfyUI · NVIDIA NIM microservices
- InferenceDynamo · vLLM · SGLang · TensorRT-LLM · Triton
- OrchestrationOxCapsule · RKE2 · Rancher · NVIDIA GPU Operator
- StorageBeeGFS · MinIO · NVIDIA GPUDirect Storage
- NetworkNVLink · Cilium · Multus · MetalLB · NVIDIA Spectrum-X
- VirtualizationKVM · NVIDIA vGPU
- Bare metal NVIDIA toolsBase Command Manager · NVIDIA Infrastructure Controller
- Token optimization
- Security
- Compliance
- Monitoring
Rack platforms for each NVIDIA generation.
32 RTX PRO 5000 Blackwell GPUs in a 20 kW, 42U, air-cooled rack. Inference, rendering and enterprise AI hosting on the tier that stands up first.
Rack-scale compute for large-model training and high-throughput inference in greenfield rooms. The volume tier of the fleet.
Next-generation capacity for the later phases of the build-out, in rooms designed for it from day one.
AM Intelligence and OXMIQ.
OXMIQ is the architecture and engineering partner for AM Intelligence Labs and its 2 GW AI compute platform.
Read the release →
AM Intelligence Labs, part of the AM Green Group, joins the round as an investor.
Read the announcement →
About 9,000 NVIDIA Rubin GPUs in Vera Rubin NVL72 rack-scale systems, with delivery in Q1 2027. The site is the first phase of a planned 1 GW of Compute-as-a-Service capacity.
Read the release →Bring us the site.
We build the AI factory.