Decoder. plain-English AI glossary

Vertical scaling

● Core

Making a single machine more powerful (more CPU, more GPU, more memory) instead of adding more machines.

Think of it like

Buying a bigger truck instead of buying more trucks.

Example

Instead of 5 GPU servers, buy 1 with 5 GPUs (or 1 huge GPU).

How it actually works

Vertical scaling hits limits — you can't make a machine infinitely powerful. It's also wasteful if you don't use the full capacity. But for certain workloads (like very large model training), you have no choice. Vertical scaling is usually simpler operationally (fewer machines to manage) but more expensive per unit of capacity.

For product teams

Vertical scaling works for bursty, high-powered jobs; horizontal is better for continuous services.

For engineers

Prefer horizontal scaling; use vertical only when you hit architectural limits.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome