The Great Inversion - When Used Drives Beat New

Executive Summary

For most of the history of hard drives, a used one was worth less than a new one and the discount was the whole proposition. You accepted unknown hours and a shorter warranty, and in exchange you paid meaningfully less per terabyte. That relationship held so reliably that it barely needed stating.

It no longer holds cleanly. Across the capacities where we currently track both a new and a refurbished offer in stock at the same time, refurbished is cheaper at most of them - the live count is in the block above, computed at page load rather than written into this sentence. That is not the finding people expect - the surprise is not that used is cheaper, it is how far the gap has widened and how erratic it has become. At some capacities the refurbished price is a fraction of the new one. At a minority of them new is actually cheaper than refurbished, which should not happen in a rational market and is itself evidence of what has gone wrong.

This report presents the spread per capacity from live data, the mechanics that produced it, and - because this is the question that actually matters to a buyer - what would have to change for it to normalise.

The spread, per capacity

The table below is computed at page load from our live catalogue. It is deliberately like-for-like: a capacity only appears if we have both a new offer and a refurbished offer in stock at that moment. Comparing a live new price against a refurbished price that sold out last week would manufacture a spread rather than measure one.

Read it with the count in mind rather than the average. A single capacity where a thin listing distorts the figure tells you little; a clear majority pointing the same way tells you something about the market.

The deepest discounts cluster in the middle of the range rather than at the frontier. That is the opposite of what you would expect if this were simply about depreciation, and it is the first clue to the mechanism.

Where new still wins, and why that matters

A minority of capacity points in our current data show new cheaper than refurbished. It would be easy to leave those out - the report reads more cleanly without them - and doing so would be dishonest, because the exceptions carry information.

A refurbished drive priced above a new one is a listing that has not repriced. The secondary market does not have a single price-setting mechanism the way a manufacturer does; it is thousands of sellers, each pricing against what they paid and what they think the market will bear, updating at wildly different speeds. When the new-drive price moves quickly, refurbished listings lag, and for a while the two cross over.

That is what an inversion looks like from the inside: not a clean new equilibrium, but a market where the ordering between two tiers is no longer dependable and has to be checked per purchase. The practical instruction is unglamorous and important - check both tiers every time, because the answer changes by capacity and by week.

The mechanics: how a shortage inverts a resale market

The chain has three links and each is documented rather than inferred.

First, hyperscaler absorption. TrendForce reported on September 15, 2025 that cloud providers were absorbing nearline HDD supply well ahead of the general market, with lead times extending accordingly. Tom's Hardware, citing DigiTimes on November 9, 2025, described enterprise HDD orders queuing beyond a year. When the largest buyers commit an entire production run in advance, everyone else is not paying a higher price for the same access - they are competing for what is left over.

Second, retail starvation. Data Center Dynamics reported on June 18, 2026 on the continuing squeeze, and Network World's January 27, 2026 coverage described the same dynamic reaching enterprise buyers who had previously never thought about drive availability at all. Retail is the residual channel: it gets what the contracted buyers do not take. In a shortage that residual shrinks first and hardest.

Third, secondary-market repricing. This is the link that produces the inversion. A refurbished drive's price is set by what the secondary market will bear, and when new supply is constrained, demand spills into the secondary market and lifts it. But it lifts it unevenly and with a lag, because the supply of refurbished drives is set by decommissioning cycles that have nothing to do with today's demand. The result is a market where the two tiers move on different clocks - and periodically cross.

What the spread over time shows

The chart above plots the cheapest live price per terabyte for each condition track over our price-history window. It is a levels chart rather than a change index: it shows where the floor sat on each date, not a percentage movement.

That distinction matters and we are explicit about it because the two are routinely confused. A percentage change computed by comparing medians across dates measures catalogue churn as much as price movement - add thirty drives to one side and the median moves without any seller changing anything. Our percentage figures elsewhere on this site use a matched-model index for exactly that reason. This chart avoids the problem differently, by plotting a floor rather than a middle, and by being honest that a floor can move because a single cheap listing appeared or vanished.

What it is good for is showing whether the gap between the tracks is widening, narrowing or noisy. Read the shape, not any single day.

Who wins and who loses

The obvious winner is the buyer willing to take refurbished at the capacities where the discount is deep. That has always been a good trade for a well-managed array with real redundancy, and it is currently a better one than usual.

The less obvious winner is anyone sitting on an installed fleet. A drive that would have been written off as having negligible resale value now has a real number attached to it, which changes the arithmetic of a capacity upgrade: the drives you are removing part-fund the drives you are installing. Our used drive value pages price that side, and the upgrade economics cluster works through whether replacing beats adding.

The loser is the buyer who genuinely needs warranty coverage. A refurbished drive's warranty is shorter and the terms vary by seller, and no discount fixes that if the requirement is contractual. For those buyers the inversion is not an opportunity, it is simply a higher price for the thing they are allowed to buy. The same applies to the RMA-versus-replace decision: when replacement drives are scarce and expensive, a warranty claim that takes weeks is a different proposition than one that takes days.

How long can this last

We publish no price forecast here, and the reason is methodological rather than coy. Our forecast page sets out a method and its error bars; repeating a prediction in a report without that machinery would be presenting a guess with a report's authority.

What we can do is state the falsifiers - the specific things that would end this condition, so you can watch for them rather than waiting for someone to announce a turn.

The inversion ends when new-drive supply reaches the retail channel again. Concretely: when the lead times described by Tom's Hardware and DigiTimes in November 2025 shorten materially, when the hyperscaler absorption TrendForce documented in September 2025 slows, or when added capacity reaches volume. The Avnet 2026 market brief, citing WSJ, IDC and TrendForce, frames the demand side as structural rather than a spike - which argues for persistence rather than a quick correction, though it is a framing rather than a date.

The inversion also ends, less happily, if the secondary market catches up. Refurbished pricing lags; lag is not permanence. If decommissioning slows while demand stays high, the used tier reprices upward and the discount compresses from the other direction.

Watch the falsifiers, not the commentary.

How to act on this

Three practical positions follow from the data rather than from opinion.

If you are buying capacity now, check both condition tracks at your target capacity every time, because the ordering is no longer dependable. The table on this page does that check live. Do not assume the tier that was cheaper last month still is.

If you are upgrading an existing array, price the drives you are removing before you buy the ones going in. That number is materially higher than it was and it changes the comparison between adding capacity and replacing it.

If you need warranty length, buy new and accept that the shortage is charging you for the privilege. There is no clever route around a contractual requirement, and a refurbished drive with a short warranty is not a cheaper version of a new one - it is a different product with a different risk profile.

Across all three: recording technology matters more than condition for a parity array. A refurbished conventionally recorded drive is a better array member than a new shingled one, and that ordering does not change with the market.