Storage Price Forecast - August 2026 Check-In

What this is, and why it exists

This is the first of a monthly series. Once a month we review the three price forecasts we publish - hard drives, SSDs and server RAM - and state what our own data did over the previous thirty days and whether the read changed.

The editions are archived rather than overwritten. That is deliberate: a forecast you can only read in its current form is not a forecast anybody can hold you to. Keeping every edition means the record of what we said, and when, stays checkable - including the months the read turns out to have been wrong.

This one carries an unusual amount of methodology, because the first thing our own numbers did this month was demonstrate a measurement trap clearly enough that publishing the naive figures would have misled everybody reading them.

The trap: two of three headline numbers had the wrong sign

The obvious way to report a month of price movement is to take the cheapest listing today, take the cheapest listing thirty days ago, and compute the change. Here is what that produces for the window 13 July to 12 August 2026:

Hard drives, floor: $11.33/TB to $12.50/TB, +10.3%. Server RAM, floor: $5.155/GB to $4.312/GB, -16.4%. SSDs, floor: $21.16/TB to $20.91/TB, -1.2%.

Two of those three are artefacts, and the reason is in the same tables. Over that window the number of listings we track went from 152 to 355 in hard drives, from 7 to 51 in memory, and from 51 to 68 in SSDs. We spent the period expanding the catalogue substantially, and a floor is the minimum of a set - add cheaper members to the set and the minimum falls without a single seller changing a single price.

The memory figure is the clearest case. A 16% "fall" computed across a window in which the tracked module count went from seven to fifty-one is not measuring the memory market. It is measuring us adding modules.

What actually moved, on the matched index

The instrument that survives catalogue growth is a matched-model index: compare every listing only against its own prior price, and average those changes. A listing that did not exist thirty days ago contributes nothing, so expanding the catalogue cannot move the number.

On that basis, for the 31 days to 12 August 2026:

Hard drives: +6.18% across 521 matched listings, of which 118 rose and 68 fell. Server RAM: +0.39% across 54 matched listings, 15 up and 9 down. SSDs: +31.96% across 72 matched listings, 12 up and 6 down.

Compare those against the naive figures above. Memory flipped from -16.4% to essentially flat. SSDs flipped from -1.2% to a large rise. Only hard drives kept their sign, and even there the magnitude came down.

One caveat we will not bury: the SSD figure rests on 72 matched listings of which only 18 actually moved, so a handful of large individual increases dominate it. It is a real signal about the listings we track and a thin basis for a claim about the whole flash market. We would rather tell you the sample size than round it into confidence.

Did the outlook change? Hard drives: no

The hard drive forecast reads that the conditions which raised prices have not reversed, and nothing this month contradicts that. The matched index is up 6.18% with roughly twice as many listings rising as falling, which is the behaviour of a market under sustained pressure rather than one turning.

The falsifiers on that page are lead times shortening in reported figures, a sustained fall in our own floor across weeks, in-stock breadth recovering, and any manufacturer coming off allocation for nearline capacity. None fired this month.

Our in-stock breadth did grow substantially, but that was our own catalogue expansion rather than the market loosening, which is exactly the distinction the section above is about. We are not going to count our own seeding as a supply signal.

Did the outlook change? SSDs: no, and the uncertainty is confirmed

The SSD forecast says the direction of underlying pressure is upward and the direction of the retail floor over the next quarter is unclear. This month's data supports both halves of that, which is a slightly awkward thing to report and the honest one.

The matched index rose sharply while the naive floor was flat to slightly down. That is precisely the pattern you would expect in a category where individual listings are being repriced upward while promotional activity keeps the visible cheap end roughly where it was. It is also why that page prints unclear rather than a direction for the floor.

The falsifiers there - QLC allocation opening up, announced supplier increases not reaching retail, our floor establishing a sustained lower level, flash and drives converging again - none fired. The read stands as written.

Did the outlook change? Server RAM: no, but watch this one

The server RAM forecast carries the strongest directional read of the three, on the basis that three manufacturers control over 95% of DRAM output, HBM took a rising share of wafers, that HBM earns several times the revenue per wafer, and 2026 HBM production is committed.

Our matched index for the month is +0.39% - essentially flat - across 54 matched listings. That is not a contradiction of the outlook, because the outlook is about sustained pressure rather than month-to-month movement, but it is the sort of result worth noting rather than explaining away. A flat month in a category we describe as under the clearest pressure of the three is a data point against our own confidence.

It is also a thin sample. Our memory history is the shortest of the three and the matched pool is the smallest. We will watch whether that flatness persists, and if it does across several editions, the read on that page needs revisiting rather than defending.

Developments in our cited sources

No new developments in our cited sources this month.

That sentence is doing real work, so it is worth explaining. Our forecasts run on a locked list of supply-side facts - the ones already published and attributed elsewhere on this site. We do not add external figures between editions, because a forecast that absorbs whatever industry commentary appeared that week is not a method, it is a feed.

When something on that list genuinely changes - a lead time figure revised, an allocation ending, a capacity commitment altered - it will appear here with the same attribution it carries on the category page. This month, nothing on the list moved.

What a buyer should take from this edition

If you are buying hard drives, the picture is unchanged and the pressure is real: the matched index is up, the falsifiers have not fired, and in a market running on allocation the risk of waiting includes availability rather than just price.

If you are buying SSDs, treat any single day's cheap listing as an opportunity rather than a level. The gap between the flat floor and the rising matched index this month is the clearest evidence we have that promotional pricing and underlying pricing have separated in that category.

If you are buying server memory, the outlook still says buy what you need, but this month's flat reading is a genuine note of caution about our own confidence rather than a signal to wait.

And across all three: check the live floors on the category pages rather than the figures in this article. This is a dated snapshot by design. The category pages recompute every few hours.