What a vCPU is
A vCPU is a virtual processor presented to your instance by the host machine’s hypervisor. It is the unit most providers use to describe compute capacity, and it is the number that appears in many configuration names.
More vCPUs help workloads that can genuinely run in parallel — builds, batch processing, simulation, media encoding, serving many simultaneous requests. A single-threaded program will not run faster simply because more vCPUs are available.
Memory is the other half
Compute sizing is always a pair: processors and memory. Running out of memory degrades a system far more abruptly than running out of processor time, because the operating system starts swapping or killing processes.
As a rule, size memory to the working set the application actually holds, then choose the vCPU count that keeps processing comfortable at peak.
Right-sizing
Most instances in the wild are larger than their workload requires. Measure real usage over a representative period before committing to a size, and prefer scaling out with more small instances over scaling up to one very large one where the workload allows it.


