Medians, Not Averages
Every aggregate on this page uses medians, so one GPU with a wild week cannot drag the market read. The individual moves behind the index also require at least 5 qualified sales in both snapshots.
This page condenses every tracked GPU with qualifying weekly sold-listing snapshots into one market read. Right now the used GPU market is roughly flat: the typical tracked card barely moved week over week.
The index covers 280 GPUs with current used pricing, backed by 1,581 recent qualified sales. Week-over-week movement is measured for the 87 cards that have two qualifying snapshots at least 5 days apart with 5+ sales each.
Release-year cohorts with at least 3 tracked, priced GPUs. The change column is the median week-over-week move among that cohort's cards with qualifying snapshots.
| Release year | Tracked GPUs | Median used price | Median weekly change | Cards with move data |
|---|---|---|---|---|
| 2025 | 29 | $609.99 | -0% | 8 |
| 2024 | 14 | $429.23 | -0.5% | 3 |
| 2023 | 23 | $508.16 | +1.9% | 4 |
| 2022 | 32 | $229.08 | 0% | 13 |
| 2021 | 36 | $290.11 | +0.7% | 6 |
| 2020 | 18 | $270.49 | -4.3% | 8 |
| 2019 | 21 | $134.99 | -2% | 5 |
| 2018 | 27 | $209.6 | +2.1% | 11 |
| 2017 | 30 | $74.35 | -1.6% | 16 |
| 2016 | 27 | $63.98 | +4.5% | 9 |
| 2015 | 23 | $59.1 | +1.5% | 4 |
What the numbers can and cannot tell you before you buy or sell.
Every aggregate on this page uses medians, so one GPU with a wild week cannot drag the market read. The individual moves behind the index also require at least 5 qualified sales in both snapshots.
Newer cohorts usually hold price while older cohorts decay, but the pattern is not smooth: a generation with strong VRAM or efficiency can hold value years longer than its raw performance suggests. Compare the cohort medians before assuming a card's price will keep falling.
The index describes the market, not any single card. Once the direction looks interesting, open the weekly movers pageor a specific GPU's detail page to see its own price history and sample size.