Skip to content
ENExplore Calculators

Vector Storage Cost Calculator

Work out vector storage cost instantly with clear inputs, formula shown and shareable results.

About Vector Storage Cost Calculator

Vector storage cost is payload plus index overhead, replicated. A 768-dimensional fp32 vector is 3,072 bytes and an HNSW graph adds roughly 30 percent for neighbour lists, so plan on about 4 KB per vector per replica. Because managed vector databases price on provisioned memory rather than disk, this is usually the dominant recurring line item in a RAG system.

Formula

Vector storage cost

bytes per vector = dimensions x 4 x 1.3; GB = vectors x bytes x replicas / 1e9; monthly = GB x price per GB-month

Worked example

With the default values below, Vector Storage Cost Calculator returns Monthly vector storage cost of $9.98. Change any input above and every figure here is recalculated in your browser.

Inputs used

Vectors storedminimum 1
5,000,000
Dimensionsminimum 1
768
Storage price per GB-monthminimum 0
0.25
Replicas for availabilityaccepts 1–10
2

Results

Monthly vector storage cost
$9.98
Storage consumed
39.94 GB
Bytes per vector including index
3,994 B
Annual cost
$119.81

Step by step

  1. 1Values used
  2. 2Vector storage cost
  3. 3Monthly vector storage cost= 9.98
  4. 4Storage consumed= 39.94 GB
  5. 5Bytes per vector including index= 3,994 B
  6. 6Annual cost= 119.81

Assumptions and limits

  • The result is derived only from the 4 inputs listed above. Anything not among them is not part of the calculation.
  • In particular, this calculator does not take inflation, tax, fees or charges or compounding frequency into account.
  • Figures are rounded for display to 2 decimal places, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.

How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.

Frequently Asked Questions

How can this cost be cut?

Quantise to int8 or use product quantisation, reduce dimensions with a Matryoshka-capable model, and archive cold partitions to a disk-backed tier.

Do I need two replicas?

For production availability yes, since a single node loses the index on failure and rebuilding takes time proportional to corpus size.

Related Calculators

Vector Storage Cost Calculator | Calcrivo