Vertical scaling
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
- Horizontal scaling — Adding more machines.
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