Hard Drive Price Forecast — Buy Now or Wait?

Decided to buy? Premium drives in stock right now

As of just now, 37 of 244 scored hard drives sit above their own recent typical level, 137 sit at the cheap end of their own record, and 70 are priced where they usually are. A further 1 are still building history. That distribution is the answer to “has waiting been working” - it is a count of badge states, not a price comparison.

If the verdict above lands for you, here is the actionable list — premium enterprise hard drives genuinely in stock this moment on Amazon US. The table balances new and refurbished listings, then ranks each half by price per TB. Buy links go straight to the live listing; specs link to the full drive page.

See the full ranking on cheapest per TB, or check where to buy for stock across every capacity and source.

The three categories, forecast separately

Storage stopped being one market. Hard drives, flash and server memory are made by different companies in different fabs under different constraints, and in 2026 those constraints diverged far enough that a single forecast for “storage” hides more than it explains. Each category now has its own page, built the same way this one is: live floors from our own catalogue, the supply-side facts we have cited, and a directional outlook that carries its date and its falsifiers.

Why “SSD prices” headlines mislead

There is no such thing as the SSD price, and most coverage of this market fails at that first step. Flash is allocated across segments that are priced separately and constrained differently, and a headline that averages them describes a market nobody buys in.

The clearest split is client against enterprise. A consumer NVMe drive and a datacentre SSD are made from the same NAND but sold into different contracts, and when a manufacturer decides where its wafers go, the segment that loses is not the one making the most noise. Enterprise and hyperscale buyers commit volume months ahead; the retail channel takes what is left. That is why a retail price can sit flat through a quarter in which the underlying component repriced sharply, and why it can then move all at once when the contracts reset.

Our own tracked set shows the same shape from the other end. The listings we price are the retail residual, and their behaviour is downstream of allocation decisions taken before the retail buyer entered the market. When our floor holds steady while the matched-model index climbs, that gap is the promotional layer sitting on top of a market that has already moved underneath it.

This is why our forecast is split into three category pages rather than published as one number, and why the SSD page prints “unclear” for the retail floor while stating the underlying pressure as upward. Those two statements are not in tension - they describe different segments of the same supply chain, and collapsing them into one direction would be the mistake this section exists to name.

Our NVMe availability work covers the decomposition in more detail, tracking how much of the catalogue is buyable rather than merely listed. Availability is the leading indicator in an allocated market; price is what happens after the queue has already formed.

How this page updates

The live figures on this page refresh every few hours with our price data. The outlook is a separate thing and it moves on a slower, deliberate cadence: once a month we publish a dated check-in that states what our own numbers did over the previous thirty days, whether the read changed in each category, and whether anything in our cited sources developed. Those editions are archived rather than overwritten, so the record of what we said and when stays checkable - including the months we were wrong.

One note on what this page displays. The only price we ever show for a product is its current one, with the time we last checked it. The trend lines and medians here are aggregate statistics we compute across the whole market - our own measurements rather than any individual listing's price history.

What the supply side says

The forecast starts with supply, because in 2026 supply is the whole story. Four data points define the market:

• Western Digital is "pretty much sold out for calendar 2026" — CEO Irving Tan on the Q2 2026 earnings call. WD holds firm purchase orders from its top seven customers and has signed long-term agreements covering 2027 and 2028. When the largest supplier is booked out a year ahead, retail scarcity is guaranteed.
• Seagate can fill only about 50-66% of near-term demand. The second major supplier is rationing allocation, which means even buyers willing to pay cannot always get units — the definition of a hard shortage rather than a price-driven one.
• DRAM and NAND relief is not expected until 2028. Micron has signaled no near-term relief, and analysts broadly expect the memory crunch to persist through 2027. That matters for HDDs indirectly: as SSD prices climb, demand shifts back onto HDDs, tightening them further.
• ComputerBase price tracking since September 2025 shows SSDs +86.6% and HDDs +50.8% on average. This is the measured reality, not a projection — the increases already happened and are holding.

Read the underlying reporting in WD is sold out for 2026, the memory chip shortage, and hard drive prices up 50%.

The wildcard: new fab capacity, arriving earlier than assumed

Every forecast on this site carries a falsifier list, and most of those falsifiers are abstractions - “if allocation loosens”, “if lead times shorten”. This section names the concrete thing that could do it, and reports that it has already started moving.

YMTC has fast-tracked its Wuhan Phase III NAND fab into mass production in the second half of 2026, roughly a year ahead of the 2027 timeline most coverage of this shortage assumed when it was written. DigiTimes reported the acceleration on 2 February 2026. Phase 3 is reported to reach 50,000 wafers per month by 2027 and 100,000 at full capacity, against roughly 200,000 combined at YMTC's first two Wuhan fabs.

That matters because the entire flash side of the current outlook rests on capacity being tight into 2027. Jefferies Equity Research, in coverage dated 29 June 2026, models storage price increases of 40-50% in Q3 2026 and 30-40% in Q4 2026 and expects no meaningful relief until 2028, explicitly because new NAND fab capacity was not expected until late 2027. If capacity arrives materially earlier than that assumption, the assumption is what breaks first.

We are not going to resolve that disagreement by averaging it, and this is the section where we say so plainly. Jefferies models relief arriving in 2028. The YMTC schedule, as reported, puts significant new NAND capacity into production a year ahead of the timeline that forecast was built on. Both are documented; they cannot both be the whole picture; and we do not have a basis for picking between them that is stronger than the sources themselves.

What would prove each side right is stated in the signals list below rather than guessed at here. For Jefferies: contract prices continuing to climb through Q4 and into 2027, which is what their model expects and what TrendForce's Q3 figure is still consistent with. For the acceleration case: YMTC output reaching the market at volume rather than merely entering mass production, and NAND contract growth stalling earlier than the 2028 horizon implies. Entering production and moving the price of a consumer drive are separated by a supply chain and, on the evidence so far, by quarters.

A wildcard is not a prediction. It is the thing we would be embarrassed not to have been watching if it lands, and it is on this page so that a reader can watch it too.

The case for waiting

In fairness, there is an argument for holding off. Some analysts suggest a possible price correction later in 2026 if hyperscaler ordering cools or if inventory that was double-booked during the panic gets released back into the channel. Seagate's HAMR ramp — the Mozaic platform pushing toward 40TB and beyond — adds real capacity per platter, and more terabytes per drive can ease $/TB pressure over time. If you have no immediate need and can tolerate the risk of the drive you want being unavailable, waiting is not irrational. But it is the weaker bet: a correction is speculative, the HAMR ramp is gradual, and neither addresses the forward contracts already locked through 2028.

The case for buying now

The stronger position is to buy at need. This shortage is powered by a multi-year AI infrastructure buildout, not a six-month chip cycle — hyperscalers are provisioning storage for training and inference workloads that will keep growing, and their forward contracts are already booked to 2028. That is fundamentally different from a shortage caused by a flood or a fab fire, where you can wait for the physical cause to resolve.

Waiting in this environment risks paying more later with no identifiable relief date to wait for. If prices do correct, the downside of having bought is modest — you paid a fair 2026 price for storage you needed. If prices keep climbing, the downside of waiting is real and open-ended. Asymmetric risk favors buying the capacity you actually require now, ideally in refurbished enterprise form for the best $/TB, and not speculating on a timing you cannot control.

Historical context — why 2026 is different

Past storage shortages resolved because their causes were finite. The 2011 Thailand floods knocked roughly 28% of global HDD manufacturing capacity offline almost overnight; prices doubled within 90 days, but the market normalized in about 18 months as factories were rebuilt and drained. The cause was physical and temporary, so recovery was a matter of time.

The 2017-18 DRAM shortage doubled memory prices over roughly 18 months and resolved in about 24 months once new fabrication capacity came online. Again, the fix was additional supply catching up to a demand level that was essentially stable.

2026 breaks the pattern. This shortage is demand-driven and multi-year: AI infrastructure is not a one-time capacity shock that supply eventually laps, but a sustained and growing consumer of storage. The historical 18-to-24-month recovery clock does not start ticking until demand growth flattens — and there is no sign of that yet. That is why the "wait it out" instinct honed in 2011 and 2017 is the wrong instinct here.

What to do

Practically: buy your current requirements now rather than deferring. Prefer refurbished enterprise drives for the best price per TB — see the live $/TB rankings and refurbished enterprise listings. For cold archive that you rarely touch, consider LTO tape, which is the one medium not spiking in 2026. And when you are ready to buy, the where-to-buy guide shows what is actually in stock right now and which alternative sources are worth your time. Check what counts as a good price per TB before you commit.

Frequently asked questions