Why Are SSD Prices Going Up in 2026? How Much of It Reached the Shelf

Ask why are SSD prices going up and you get a cause and a percentage: AI data centres are buying the NAND, suppliers cut wafer starts, prices are up as much as 60%, buy now. The cause is well sourced and the percentage is real — but it is a contract percentage, agreed between suppliers and large buyers, and not one of the four pages we read separates it from what a drive costs at a US retailer. That gap is the whole story this page tells, and we measure it rather than assert it: the published contract range for the quarter against our own observed retail index over the same period.

So the answer has two halves and almost everything published carries only the first. We read TrendForce’s own releases for the contract side, scored the four pages ranking for this question on 4 October 2026, and priced every drive we track per gigabyte with the retailer and the date attached. We make no forecast below: the only projections on this page are TrendForce’s own, labelled as theirs.

What a gigabyte of NVMe costs at US retail today

Start with the number the question is actually about. A percentage move means nothing without a level, and a level means nothing without a unit. Across the 7 tracked drives carrying a live in-stock price, a gigabyte runs from $0.1845/GB on the Crucial T500 2TB NVMe Gen4 SSD to $0.2437/GB on the Kingston KC3000 2048GB NVMe Gen4 SSD — 32.1% apart for storage that does the same job.

The median gigabyte across that basket is $0.1991/GB, or $199.13/TB, from 7 drives with current prices. That is the figure to hold onto, because it is the one every page in this SERP leaves out — and it is also the figure that makes their numbers checkable. One of the four pages publishes dated capacity ranges — a 2 TB NVMe drive at $120 to $160, as of 12 April 2026 — which sit well below anything we can observe in stock today. We are not calling it wrong: a figure dated April and a figure dated October are not the same claim, and that page names no retailer and no basket. That is the point: an undated price with no named basket cannot be compared with anything, including the price you are about to pay.

Contract prices and shelf prices are two different numbers

Nearly every percentage on this subject comes from one place: the quarterly contract-price reporting memory analysts publish for suppliers and large buyers. Those are real figures and they are not retail figures. A contract price is negotiated per bit, quarterly, in volume; a shelf price carries a retailer’s margin, inventory bought at an older cost, and whatever the listing is doing that week.

Two details in that table matter more than the headline ranges. The first is that the 4Q26 release — published Sep 30, 2026, after the quarter we measured had closed — says suppliers took a more flexible pricing stance on client SSDs, restraining the increases, while enterprise SSDs were the one category accelerating. The consumer drive and the data-centre drive are not on the same curve, and a page that quotes an enterprise number to a consumer is answering a different question. The second is the direction of travel on the memory side: conventional DRAM was published at +13–18% for 3Q26 and +10–15% for 4Q26 — still rising, by less.

Why are SSD prices going up: how much reached the shelf

Here is the measurement no competing page makes. We have a published contract range for a quarter, and our own index over a window sitting almost entirely inside it: the median price per gigabyte across every tracked drive, normalised to 100 at the series start. Setting one against the other says how much of a contract move has shown up where a buyer meets it.

The arithmetic is withheld whenever our index did not rise over the window, because a share of a rise that did not happen is not a number worth printing. The table says which case we are in.

Three things could produce a gap like that, and we can name them without weighing them: retailers are still selling inventory bought at an older cost, margin absorbs part of an increase before the shelf shows it, and a contract forecast is a projection rather than a print. None of that discredits either figure; they measure different things.

One listing carried most of the movement

An index is a median, and nobody buys a median. The question a reader actually has is whether the drive they want went up, and that is a different question with a different answer — sometimes the opposite answer.

The split is stark. Kingston KC3000 2048GB NVMe Gen4 SSD moved +39.1% — $140.35 in cash — and carried 50.5% of all the cash movement in the basket across 7 drives. Remove it and the median tracked drive moved +0.2%, against +1.0% with it included. We measure that in cash rather than percent deliberately: the index is computed on prices, so a large percentage move on a cheap drive and a small one on an expensive drive are not the same event for it.

What we will not do is explain the big one. A listing repricing far outside its peers suggests a stock or availability event at that retailer rather than a market move, and we have no primary source for the drive’s status, so the page reports the observation and stops. The consequence is the useful part: one listing can move a category index several points while the drive in your basket does nothing, and a headline percentage cannot tell you which of those you are in. A watch on the drive on the live chart can.

Storage did not move like memory

The shortage is usually told as one story about memory. Our two indexes are computed identically from the same scrapes — lowest in-stock price per SKU, divided by capacity, median across the basket — which is what makes comparing them worth anything.

The two come out close — −4.4% on storage and +7.5% on DDR5 — and the caveat matters more than the closeness: they are not the same window. Our storage index runs to Oct 21, 2026 across 23 points, while the DDR5 index’s most recent point is Aug 13, 2026 across 4 — 14% of 3Q26 against storage’s 65%. Two measurements over different periods are not one comparison, so the useful reading is narrower: each side moved by single digits at retail while the contract ranges published for them that quarter were further apart than they are — +10–15% for NAND against +13–18% for conventional DRAM. Measured the same way as the storage row and over its own shorter window, the DDR5 shelf delivered 42% to 58% of its contract range.

Why the published answers disagree with each other

We read the four pages ranking for this question on 4 October 2026 and scored them on what a buyer can check. A fifth, a vendor-forum post, could not be read by a plain fetch, so it is counted and left out rather than scored on its title.

The pattern is consistent: the cause is well covered, the price is not. None publishes a price per gigabyte, two publish no dollar figure at all, and the one dated price table is six months old at reading with neither retailer nor basket named. Nor does any of them draw the contract/retail line: the largest percentage on the page — NAND costs up 246% — sits in an article about what a buyer should budget, with no market attached. All four answer the timing question with a forecast, which may turn out right and is not checkable today.

What to do about it, without a forecast

None of this says prices are not rising. It says the number you were shown came from a different market, and that the retail move sits next to a spread you can capture today without waiting for anything.

That is the comparison worth carrying away. The 2 TB band runs 35.2% from cheapest to dearest right now — $130.00 on the same amount of storage — which is 8.0x the move the whole category made across our window. Picking the drive is the larger decision by a wide margin, and it needs no view on what happens next.

Price the gigabyte, not the drive. Treat an undated price as no price, including ours. Decide the capacity first, then compare inside it: the per-gigabyte curve by capacity is steeper than the brand ordering, as the 2 TB against 4 TB step shows. And watch the listing you intend to buy, because any headline percentage is a median of lines you are not buying.

How these numbers are produced

Every retail figure above is a dated in-stock observation at a tracked US retailer. Prices come from live product pages at Newegg and B&H, both of which permit it in robots.txt and robots.txt — we record that check before a retailer’s first scrape, and Amazon is only ever reached through the Keepa API.

An out-of-stock listing records no price rather than carrying the last one forward, and a move over 60% waits for a second scrape, so a retailer glitch cannot appear here as a market event. A basket median needs five drives, a capacity band three, a per-drive move four observations; nothing older than 30 days is shown as current. Demo rows are blocked at the database layer rather than by convention, so a fabricated price cannot reach this article.

Related reading

FAQ

Context: NVMe Gen4 from $0.184/GB, DDR5 from $16.53/GB, cheapest in-stock observations in the last 30 days.