Where the supply comes from
Enterprise drives do not end up used because they failed — the overwhelming majority are decommissioned while still healthy. Two channels dominate. First, hyperscaler and colo refresh cycles: data centres retire whole fleets on a fixed schedule (commonly three to five years) regardless of individual drive health, because the economics of rack space, power, and failure-rate risk favour swapping the fleet over nursing it. Second, manufacturer warranty-return channels: an RMA’d drive gets tested, wiped, and resold as recertified stock rather than scrapped, because the silicon and platters are worth recovering even when a single unit in a batch tested bad. Neither channel is scraping the bottom of the failure curve — see the recertified-drives guide for how those two source types map onto warranty and expected failure rate.
This is also why used enterprise supply tracks the hyperscaler capex cycle more than it tracks consumer demand: a wave of decommissions from a handful of large operators can move the used-market price of a specific capacity tier independent of anything happening in new-drive retail.
Why the discount isn’t flat across the lineup
The used discount is largest where decommission volume is heaviest and smallest at the edges of the capacity curve. Midrange enterprise capacities — the tier that made up the bulk of hyperscaler deployments a refresh cycle or two ago — are where used and recertified SAS/nearline supply is deepest, and where competition among resellers pushes the discount versus new the hardest. At the low end, small-capacity enterprise drives are rarely worth reselling at all relative to shipping and handling, so there is little used inventory and little discount to speak of. At the very top of the current capacity curve, drives are still overwhelmingly in first deployment — nothing that large has finished a decommission cycle yet — so used supply is thin and prices sit close to new. The exact boundaries of that middle band move as the industry’s capacity ceiling moves, which is why we track it live rather than writing it down as a fixed number: compare current bare-drive pricing by condition on the SAS and cheapest-per-TB rankings, both sortable by condition, to see where the gap is widest this week.
Interface matters as much as capacity. SAS enterprise drives see a proportionally deeper used discount than SATA nearline at equivalent capacity, mostly because SAS has a smaller buyer pool outside data-centre and homelab use — less consumer demand competing for the same used supply keeps prices softer.
Reading a seller before you read the listing
On eBay specifically, the seller tells you more than the title does. A few signals worth checking before price:
- Feedback volume and specialisation. A seller with thousands of transactions specifically in server hardware or data-centre pulls has a testing pipeline; a seller with broad general-merchandise feedback and a handful of drive listings usually does not. Volume in the specific category is the signal, not overall feedback percentage.
- Return window length. 30 days or more suggests the seller is confident enough in their testing to eat an infant-mortality failure; a 14-day or no-returns listing pushes that risk onto you. Weigh this the same way the recertified-drives guide treats seller-refurbished warranty claims — price the listing as if the stated return window is the only protection you get, because it is.
- Listing specificity. A listing that states power-on hours, SMART pass/fail, and the exact pull source (e.g. “pulled from Equinix decommission, tested and wiped”) is a seller who tested the drive. “Pulled from working environment” with no SMART data and a stock manufacturer photo is a seller who did not.
- Lot listings vs single-drive. Bulk lots are usually priced for the reseller’s margin, not tested per-drive — expect a higher dead-on-arrival rate and budget for it, especially if you are buying to fill a large array rather than one bay.
What this means for buying decisions
The practical read: if you are filling an array in the capacity tier where used supply is deep, the discount is usually real and worth taking, provided the seller signals above check out. If you are chasing the newest, largest capacity point, expect to pay close to new-drive prices used or skip the used market entirely for that capacity — the supply simply is not there yet. Run the actual numbers for your case, not the general pattern, with the used-vs-new calculator, which takes the live price gap at your chosen capacity and interface, together with the failure-rate assumptions from the recertified-drives guide, and tells you the exact price above which the used discount stops being real.