Solving for Minimum Efficient Object Size
We want to find the object size at which savings exactly equal the monitoring fee — the break-even point. Below this size, S3-INT costs you money. Above it, S3-INT saves you money.
Starting from the inequality:
$$\text{Size in GB} \times \$0.0105 = \$0.0000025$$
Solve for size:
$$\text{Size in GB} = \frac{\$0.0000025}{\$0.0105} \approx 0.000238 \text{ GB}$$
Convert to kilobytes:
$$0.000238 \text{ GB} \times 1{,}024 \times 1{,}024 \approx \mathbf{250 \text{ KB}}$$
The Rule of Thumb
Objects must be larger than ~250 KB for S3 Intelligent-Tiering to be financially beneficial.
Below 250 KB, you pay more in monitoring fees than you save in storage costs. This is true even if the object perfectly matches the tiering pattern (uploaded once, never accessed again).
| Object Type | Typical Size | S3-INT Outcome |
|---|---|---|
| High-res product image | 2–5 MB | Strong net gain |
| PDF report | 500 KB–2 MB | Net gain |
| Compressed log chunk | 200–400 KB | Marginal — verify |
| Raw log line | 1–10 KB | Net loss (also below 128 KB threshold) |
| Thumbnail image | 20–80 KB | Net loss (also below 128 KB threshold) |
The Archive Instant Access Tier
S3-INT also includes an Archive Instant Access tier for objects not accessed for 90 days. The storage cost drops to ~$0.004/GB, increasing the spread to ~$0.019/GB. This lowers the break-even point to approximately 130 KB — still above the 128 KB exemption threshold, but only barely. For objects between 128 KB and 250 KB, the Archive Instant Access tier is what makes S3-INT marginally viable, but the margin is thin enough that you should model it explicitly rather than assume savings.