Glossary

The AI compute terms new developers search before they trust the route

Use this glossary to learn the key ideas behind AI compute without getting pushed straight into hardware hype. Each term is defined in plain language, then tied back to workload execution and what matters in practice.

Open the beginner hub Read the FAQ

Beginner definitions

Best for

Good for first-time builders who need vocabulary before deeper guide pages.

Workload-first meanings

Focus

Terms are explained through what they mean for execution, not just raw hardware.

Branch into guides

Next move

Use the term pages as bridges into planning, pricing, and product pages.

Why this glossary exists

Definitions should reduce confusion, not create a shopping list

Many AI compute definitions online start with hardware jargon and assume the reader already knows what kind of workload they are running. That is backwards for a beginner. This glossary explains the term, then explains why it matters for real AI workloads.

The goal is not to turn every definition into manual infrastructure advice. The goal is to make the concepts clear enough that a developer can ask better questions about fit, cost, and execution.

Glossary terms

Start with the concept you are stuck on, then follow the related links into guides and product pages.

what is a gpuGPUA GPU is a processor designed for highly parallel computation. In AI, GPUs matter because they make model inference and training much faster than general-purpose CPUs for many workloads.what is vramVRAMVRAM is the memory attached to a GPU. In AI workloads, VRAM matters because it affects whether a model can fit and run cleanly on a given route.what is latencyLatencyLatency is the time it takes for a system to respond to a request. In AI apps, latency matters because it shapes whether the user experience feels instant, acceptable, or frustrating.