Graphics Processing Unit (GPU)
The short answer · 11 words
In short
A GPU is a parallel processor for many similar computing operations.
Flow-alerta.cab (2026): Graphics Processing Unit (GPU). https://flow-alerta.cab/en/glossar/graphics-processing-unit-gpu. Editorially reviewed on: 10 October 2026.
How much detail would you like?
Practical
GPUs enable modern AI workloads, but they cause acquisition, energy and operational expenses.
Hypothetical example: A company compares local GPU inference with cloud API based on latency, utilization and overall costs.
Compare your actual workload by memory, runtime, utilization and total cost.
At a glance
- Core
- Many computing units process vector and matrix operations simultaneously. This accelerates graphics, ML training and inference, as long as software and data flow fit.
- Use
- GPUs enable modern AI workloads, but they cause acquisition, energy and operational expenses.
- Example
- Hypothetical example: A company compares local GPU inference with cloud API based on latency, utilization and overall costs.
- Limits
- Equate VRAM with computing power; ignore cooling, drivers, data types and utilization; just compare list price.
Questions about this term
Does every AI need a GPU?
No, the need depends on model, task and software.
What is VRAM?
Graphic memory; for integrated GPUs, memory may be shared with the CPU.
All details and background
Many computing units process vector and matrix operations simultaneously. This accelerates graphics, ML training and inference, as long as software and data flow fit.
ELI5 – Explained simply
Many small computing forces work on similar tasks in parallel.
Why it matters in practice
GPUs enable modern AI workloads, but they cause acquisition, energy and operational expenses.
Example
Hypothetical example: A company compares local GPU inference with cloud API based on latency, utilization and overall costs.
Pitfalls and limits
- Equate VRAM with computing power; ignore cooling, drivers, data types and utilization; just compare list price.
Origin & history
GPUs are used for graphics processing and are also used for general parallel calculation; NVIDIA documents the CUDA programming model for this purpose.
Use in mid-sized companies
- Compare your actual workload by memory, runtime, utilization and total cost.
Related terms
- CUDA
- VRAM
- TPU
- Accelerators
- Inference
Matching Flow-alerta service
AI opportunity analysis & roadmapAI consulting: Examine tasks, data and prerequisites of a possible AI use case.Matching tool / use case
View use casesNVIDIA CUDA for supported GPUs; measurement of memory requirements and utilization.Sources
Quellenprüfung: 10.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.
Editorially reviewed on: