Learn
AI cloud, explained.
Short, sourced answers to the questions teams ask before they build an AI cloud of their own.
What is a neocloud?
A neocloud is a cloud provider built around GPU capacity for AI. What sets neoclouds apart from hyperscalers, who they are, and the software layers they run.
What is GPU as a service (GPUaaS)?
GPU as a service rents GPU compute on demand. How GPUaaS is priced, what customers expect, and what an operator needs to launch a GPUaaS offering.
What is GPU cloud management software?
GPU cloud management software turns GPU servers and clusters into a multi-tenant cloud. The capabilities that matter, build vs buy, and questions to ask vendors.
What is a sovereign AI cloud?
A sovereign AI cloud keeps data, models, operations and control under the jurisdiction you choose. What sovereignty requires in practice and how to build one.
Sovereign cloud vs private cloud: what is the difference?
A private cloud is dedicated infrastructure; a sovereign cloud adds jurisdiction, operator and control-plane independence. A side-by-side comparison.
What is an AI factory?
An AI factory is data center infrastructure built to produce AI at scale: training models and serving inference. Its layers, and the software that runs it as a service.
What is air-gapped AI, and how do you deploy it?
Air-gapped AI runs models and the platform around them on networks with no internet connection. What truly air-gapped means, and the steps to deploy it.
How does GPU multi-tenancy work on Kubernetes?
How to share GPUs between tenants on Kubernetes: hardware partitioning, time-slicing, vGPU and dedicated nodes, plus namespace, virtual cluster and cluster isolation.
How do you migrate VMware VMs to KubeVirt?
Moving virtual machines from VMware vSphere to KubeVirt on Kubernetes: the tools, a step-by-step plan, and the issues that slow migrations down.
How do you run LLMs on premises?
Running large language models on your own hardware: reference architecture, choosing between vLLM, SGLang, NVIDIA NIM and Ollama, sizing GPUs, and serving many teams.
What is FinOps for AI, and how do you meter GPU usage?
FinOps for AI applies cloud financial management to GPUs and models: what to meter, showback vs chargeback, price books, and keeping GPUs utilized.
Slurm vs Kubernetes for AI: which should you use?
Slurm and Kubernetes both schedule AI workloads. How they differ, when each fits, and how teams run both on one GPU estate.