HDD Price Index: Dated $/TB for 8TB NAS Drives from Direct US Retailers (Pilot)

By LK Wood IV · Published 2026-08-12 · ~5 min read · St. Louis County, MO

Capture paused on 2026-09-14. I stopped capturing this index on September 14, 2026. The last capture ran on September 7, 2026, and the published observations end there. No capture ran from September 8 through 12, and the September 13 attempt failed before it wrote any rows. In 17 captures the pilot never reached its six-line gate. The most any capture accepted was four lines, so no index row was ever written. Everything below describes captures up to September 7.

The short version

Consumer NAS drive prices rose roughly 30-50% through the 2026 supply squeeze as WD and Seagate steered nearline output toward AI and cloud buyers, and most of the $/TB numbers that circulate are undated screenshots or grey-market recertified listings quoted as if they were new-drive prices.

This page starts a dated, checkable record: a $/TB index for 8TB 3.5-inch internal drives — the mainstream homelab size — observed at named direct US retailers only, with the method, the acceptance rules, every rejection reason, and the raw data published alongside the numbers.

It is labeled a pilot because the index metric has an activation gate it has not passed yet. Day one produced exactly one qualifying observation, and rather than loosen the rules to manufacture a fuller table, the gate status is printed here verbatim.

Day-one observations (2026-08-12)

1 qualifying offer across 1 drive line, out of 673 listings reviewed for 12 tracked models (four NAS, two surveillance, two enterprise, three desktop lines, plus IronWolf Pro). For scale: camelcamelcamel’s tracking of this exact SKU records a $189.99 Amazon low in July 2024, about $23.75/TB, so the day-one qualifying offer sits at roughly 2.1x that tracked 2024 low.

Gate status, verbatim: 3 of 12 tracked drives produced qualifying offers on 2026-09-07, from 607 listings reviewed; the index metric (median cheapest $/TB across drives) activates at 6, so it stays closed and no index row was written.

The capture’s blocking reasons: “Direct retailer stock status is not independently confirmed.”; “Exact seller capacity/variant must be verified on the retailer page before publication.”; “Only 3 of 12 models produced eligible rows; the minimum is 6.” Every published row carries verification_status: discovery_only.

The NAS drive lasted exactly one capture. The September 4 capture recorded the finding that this index, named for 8TB NAS drives, held none of them: BarraCuda and WD Blue are desktop drives, WD Purple is surveillance, and IronWolf, WD Red Pro, Exos 7E10 and WD Gold had all stopped qualifying.

Seagate IronWolf 8TB qualified again on September 6, and that day’s entry recorded the segment gap as closed. It did not stay closed. IronWolf produced no qualifying row on September 7. The basket holds no NAS drive again.

An enterprise line replaced it, so the count held at 3 while a third of the basket turned over. Seagate Exos 7E10 8TB (ST8000NM017B) entered at $408.41, or $51.05 per TB, at Best Buy.

Exos 7E10 was one of the four lines the September 4 entry named as having stopped qualifying. It returns after that gap. A reader watching only the headline would see 3 of 12 on both days and conclude nothing changed. The count held. The basket did not.

It came back 45 percent dearer, then left again without repricing. IronWolf 8TB sat at $329.99 ($41.25/TB) on August 24, August 25, August 31 and September

1. Four captures across nine days, flat to the cent. On September 6 it was $479.17, or $59.90 per TB. The August 31 and September 1 listings name part ST8000VN004, and so did the September 6 listing. That reading was like-for-like on the manufacturer’s own identifier. It remains the largest move this page has recorded. On September 7 it produced no qualifying row at all. It did not fall back. It stopped surfacing.

One confound, named rather than buried. Those part-confirmed $329.99 rows were at Newegg. The $479.17 was at Best Buy, so a retailer swap ran alongside the price move. The sibling DDR5 index on this site has now shown three separate times that a swap alone can register as a double-digit move which later round-trips.

Best Buy did list an 8TB IronWolf at $329.99 on August 25, which points the same way, but that title never stated the part. The jump is recorded and its caveat is recorded beside it. One capture did not settle it, and the line has now left the basket without a second reading.

Both desktop lines held, to the cent and at the same seller. Seagate BarraCuda 8TB has been $249.99 ($31.25/TB) at Best Buy for four consecutive captures, September 4 through September 7. WD Blue 8TB (WD80EAAZ) has been $324.99 ($40.62/TB) at Best Buy for five, since September 2. The cheap end has not moved.

WD Purple left without repricing. WD Purple 8TB (WD85PURZ, $435.01, $54.38/TB) qualified on September 5 and has not surfaced since.

All three accepted offers sit at one retailer. Best Buy supplied every qualifying row on September 7, and on every capture since September 2. That is five captures in a row from one shopfront. A median built from a single shopfront would measure that shopfront rather than the market. That is a further reason the six-line minimum exists.

That is the whole argument for the gate. Three of twelve lines is a poor basis for a market index, a headline count is a poor measure of stability, and a basket can drift out of its own stated segment while the count never moves. The gate exists to say so out loud instead of averaging any three drives into a number.

What gets rejected, and why that’s the point

604 of the 607 listings reviewed on 2026-09-07 were rejected, each with recorded machine-readable reasons (a listing can carry several): 529 came from sellers off the direct-retailer allowlist — eBay storefronts, recertified-drive houses, white-label resellers, marketplace suffixes; 422 never named the exact tracked model line; 210 never stated the 8TB capacity in the title; 118 carried prices outside the sanity band; 59 were multi-packs or bundles; 46 named a sibling line (IronWolf Pro under the IronWolf query, SkyHawk AI under SkyHawk); 17 were external or enclosure products; 5 were used, recertified, or open-box.

The capacity bucket is the HDD-specific finding: direct retailers frequently publish family-level Google Shopping titles — real examples from this capture include “Gold Hard Drive WD” and “Exos 7E10 Hard Drive Seagate” at $399.99 — that never say which capacity the price belongs to. A tracker that guessed would fill its table faster and be worthless. The rejection log ships with the dataset.

Method, exactly

Captures run against Google Shopping (US, English) via DataForSEO’s merchant API, one exact-model query per tracked drive.

An offer qualifies only if the seller matches a named direct-retailer allowlist (Best Buy, Newegg, B&H, Micro Center, Adorama, Walmart first-party, and peers — marketplace suffixes excluded), the title states the exact line and the 8TB capacity, no sibling-superset line token appears, the product is a single internal drive (external/enclosure listings and multi-pack markers like “2-Pack”, “2 x”, or “20PK” are rejected), condition is new (used, refurbished, recertified, and white-label all rejected), and the USD price sits inside a $100-600 band.

Two method corrections were made against measured evidence on day one, both recorded in the change log below: the price ceiling was raised from $400 after a real single-drive offer — the WD Red Pro at $402.99 — died on the old band as its only rejection reason, and a multi-pack guard was added in the same pass because pack listings ($599.98 to $1,479.96 for 2- and 4-packs in the same capture) pass every title rule and would otherwise have entered the raised band as fake singles.

A third correction landed on 2026-08-30 (method v2). “States the 8TB capacity” now includes a title carrying that line’s own declared manufacturer part number as a standalone token — ST8000VN004, WD8005FRYZ, and the ten others named in the tracked labels — and “8000 GB”, the same statement in the other unit. Refusing the part number while trusting marketing prose was backwards: the part number is the manufacturer’s own encoding.

It is very nearly not a loosening, and the exception is worth naming rather than claiming none exists: a title that states a conflicting capacity in prose while carrying the line’s declared 8TB part number now passes where v1 rejected it. No such title occurs in any of the ten stored captures, so the measured yield of the change is still zero, but the absolute claim this sentence used to make was wrong.

A truncated stem is still rejected, because HDWR names the Toshiba X300 family from 4TB to 18TB and states no capacity at all, and a part number embedded in a rebadger’s longer SKU is still rejected because that is a guess about something the manufacturer did not write.

Amazon is excluded by design (see FAQ). The full collector, product basket, tests, raw responses, and rejection logs are in the site’s repository.

Use the data (CC BY 4.0)

Download every qualifying observation this pilot recorded: hdd-price-observations.csv, one row per qualifying offer, with capture date, drive, price, $/TB, seller, and verification status, from 2026-08-12 to the last capture on 2026-09-07.

The daily index series (median cheapest $/TB across drives, one row per UTC day) was designed to publish at /data/hdd-price-index.csv once six lines qualified on one day. None did before I paused capture, so that file does not exist. License: Creative Commons Attribution 4.0. Attribute as “TechFuelHQ HDD Price Index” linking to this page.

Deciding between the drive lines this index tracks? The IronWolf vs IronWolf Pro vs Exos comparison covers workload ratings and warranty, the budget NAS build guide prices whole builds in the drive crunch, and the NAS storage calculator turns raw capacity into usable space under RAID. For why the prices look like this and what to do about it, the 2026 HDD price crisis explainer is this dataset’s narrative companion.

The zero-offer day, and what it was actually made of

Two captures ran on 2026-08-29 — 07:57 and 14:03 UTC, 704 and 671 listings — and both passed nothing at all, from four qualifying lines five days earlier under an identical method. A zero-eligible day is a legitimate result during this squeeze, and it was published as one. That is exactly why it was worth auditing instead of believing: a report of a broken instrument still reads as healthy.

Half of the drop was the instrument. Google Shopping’s headline-offer titles had moved from marketing prose to the manufacturer part number, and the collector could read a capacity only out of prose.

The clearest case is WD Gold at Central Computers, $369.99: the same seller, the same price, the same Google catalog product that qualified on 2026-08-24 as “WD Gold 8TB 3.5” Internal Hard Drive" came back titled “Western Digital WD Gold wd8005fryz Hard Drive” — and died with capacity as its only rejection reason. Seagate IronWolf did the same thing at B&H, $389, as “Seagate IronWolf ST8000VN004 Hard Drive”. Method v2 reads both.

Half of it was the market, and stands. The Exos 7E10 headline offer left Adorama for “Newegg.com - PC Parts Hub” at $382 — a marketplace suffix — and WD Red Pro left Newegg for AVLgear.com at $455. Neither seller is on the direct-retailer allowlist, and both rejections are correct. Rows from allowlisted retailers fell from 102 on 08-24 to 77 and 63 on 08-29: real thinning, not a collapse.

And most of what remains capacity-blocked on those days is still genuinely ambiguous and still correctly rejected — “WD Blue Hard Drive” at Best Buy ($324.99), “Hard Drive WD Purple” ($427.07), “Exos 7E10 Hard Drive Seagate” at Newegg ($349.99), “Toshiba X300 Hard Drive HDWR” at Micro Center ($385.99). Those are real offers this index refuses to guess at.

Replayed against every stored capture, v2 changes nothing on the eight days from 2026-08-12 to 2026-08-25, and turns the two 2026-08-29 zeroes into two lines and one line. Six are needed. The gate stays closed, and no row is published from this replay — the numbers above come from stored evidence, not from a capture.

At the time of this 2026-08-30 replay, the downloadable file was unchanged at 2026-08-25. When the next live capture runs under v2, its first rows will be partly method-attributable rather than a pure market move, and the series file’s method_version column marks them as such.

Every qualifying observation

The file below holds one row per qualifying offer for each retained capture date, from 2026-08-12 to 2026-09-07. It will not gain rows while capture is paused. Its newest rows come from the September 7 capture, which accepted three offers across three tracked lines out of 607 listings reviewed, all at Best Buy.

Seagate BarraCuda 8TB held at $249.99 ($31.25/TB), WD Blue 8TB held at $324.99 ($40.62/TB), and Seagate Exos 7E10 8TB entered at $408.41 ($51.05/TB). No NAS line qualified, so the six-line gate stayed closed and no index row was written.

The 2026-08-25 captures produced this pilot’s clearest churn lesson. Its first capture surfaced the WD Red Pro cluster at Best Buy for $489.11, a valid same-listing move against August 23’s $402.99 (docid and seller both match). Its third capture surfaced the same docid at Newegg for $379.99.

Twelve minutes apart. $109.12 apart. A single capture is one draw from a churning listing pool, which is why this page publishes a dated series with its method instead of a screenshot. The table renders from the raw CSV at build time, so it always matches the downloadable file.

The day-one table above is kept as captured on 2026-08-12. It is not rewritten.

Rendered at build time from the raw CSV — the table and the download can never disagree.

Change log

• Capture paused (2026-09-14). No observation changed, so the dataset stays at 0.9.11. The last capture ran on 2026-09-07. No capture ran from 2026-09-08 through 2026-09-12, the 2026-09-13 attempt failed without writing rows, and HDD capture was switched off on 2026-09-14. Across 17 captures on 14 dates the basket peaked at four lines, so hdd-price-index.csv was never written. Corrected in the same pass: the FAQ called captures daily and named 2026-09-06 as the newest capture, the observations section described the 2026-09-04 capture under a 2026-09-06 date, and the 0.9.11 entry below said Best Buy had supplied every row for three consecutive captures when it was five.

• 0.9.11 (2026-09-07) — 607 listings reviewed, 3 eligible offers across 3 of 12 lines, gate blocked, no index row written and hdd-price-index.csv still never written. The count held at 3 while a third of the basket turned over: Seagate IronWolf left after a single capture, without repricing, and Seagate Exos 7E10 8TB (ST8000NM017B) entered at $408.41 / $51.05 per TB. The accepted basket holds no NAS drive again, so the September 6 entry’s “segment gap closed” reading is bounded to that one capture. Both desktop lines held to the cent at Best Buy, BarraCuda a fourth capture and WD Blue a fifth. All three accepted rows came from Best Buy for a fifth consecutive capture (corrected 2026-09-14; this entry first said third).

• 0.9.10 (2026-09-06) — 645 listings reviewed, 3 eligible offers across 3 of 12 tracked lines, gate closed at half the minimum. Seagate IronWolf 8TB qualified again, the first NAS drive in four captures, closing the segment gap this page recorded on September 4. It returned at $479.17 ($59.90/TB) against the $329.99 ($41.25/TB) it held to the cent across four captures from August 24 to September 1, up 45.2 percent on the same part ST8000VN004. The retailer changed from Newegg to Best Buy in the same step, so the confound is recorded beside the jump. Both desktop lines held to the cent at the same seller: BarraCuda $249.99 for a third capture, WD Blue $324.99 for a fourth. WD Purple left without repricing. All three accepted rows sit at Best Buy. This release also resyncs the 2026-09-05 capture, whose rows shipped without a prose update because step 6b of the capture task ties the resync to an appended series row, which a gate-blocked day never produces.

• 0.9.9 (2026-09-04) — 680 listings reviewed, 2 eligible offers across 2 of 12 lines, gate closed for a fourth capture running. No capture ran on 2026-09-03. WD Blue 8TB held $324.99 at Best Buy to the cent, the first line to survive a capture boundary since the IronWolf and WD Gold streak ended. WD Purple left; Seagate BarraCuda 8TB entered at $249.99 / $31.25 per TB, its first appearance in the series and 23 percent below the previous cheapest accepted figure. Flagged rather than headlined: it is a 5400 RPM desktop drive, and no NAS drive qualified at all, so the accepted basket has drifted out of the segment this index is named for. No index row written.

• 0.9.8 (2026-09-02) — 690 listings reviewed, 2 eligible offers across 2 of 12 lines, gate closed for a third consecutive capture. The count is unchanged and the basket is entirely new: IronWolf 8TB and WD Gold 8TB both left the result pool without repricing, ending the eight-day identical-price streak by absence rather than by a price move, and WD Blue 8TB ($324.99, $40.62/TB) and WD Purple 8TB ($435.01, $54.38/TB) entered for the first time in the series, both at Best Buy. The turnover crossed segments, NAS and enterprise out, desktop and surveillance in. Rejection counts re-derived from this capture.

• 0.9.6 (2026-08-31) — first capture under method v2, and it confirms the 0.9.5 diagnosis rather than merely applying it. 699 listings reviewed, 2 eligible offers across 2 of 12 lines, the first non-zero capture since 2026-08-25. The two accepted lines are precisely the two v2 was built to recover, Seagate IronWolf 8TB and WD Gold 8TB. Both sit at the same price and seller they held on 08-24 and 08-25. The recovery is instrument repair, with no repricing behind it. The two lines diagnosed as genuine market moves, Exos 7E10 and WD Red Pro, remain correctly rejected. The 4 to 0 collapse is now fully accounted for, half instrument and half market, as 0.9.5 predicted. Gate stays closed at 2 of 6 and no index row was written.

• 0.9.5 (2026-08-30) — method v1 → v2, from a diagnosis rather than a capture. Two captures on 2026-08-29 (704 and 671 listings) each passed zero offers across zero lines, against four lines on 2026-08-24 under an unchanged method, so the step change was audited row by row. Two of the four lost lines were a collector defect: Google Shopping retitled direct-retailer offers from prose to the manufacturer part number, and v1 could read a capacity only out of prose, so it rejected “Western Digital WD Gold wd8005fryz Hard Drive” ($369.99, Central Computers) — the same seller, price, and catalog product it had accepted on 08-24 — with capacity as the only blocker, and “Seagate IronWolf ST8000VN004 Hard Drive” ($389, B&H) likewise. Same class as the RAM index’s 2026-08-29 fix. The other two were genuine: the Exos 7E10 and WD Red Pro headline offers moved to sellers off the allowlist (“Newegg.com - PC Parts Hub” $382, AVLgear.com $455), and those rejections are correct. v2 accepts a title carrying that line’s own declared 8TB part number as a standalone token, plus “8000 GB” (measured yield across all ten stored captures: zero offers; it closes a blind spot, it does not move a number). Basket v1.1 declares the twelve part numbers machine-readably and the loader refuses any that the human-authored label does not already name. Refused on purpose, with a real offer as the cost: the truncated stem “Toshiba X300 Hard Drive HDWR” (Micro Center, $385.99) stays rejected because HDWR spans 4TB to 18TB. Nine new tests, five of which fail against v1; 28 pass. (An earlier version of this entry said “six new tests, all failing against v1” - both numbers were wrong, corrected 2026-08-30 by counting the suite at 19 -> 28 and re-running it against the v1 module.) Replayed against all ten stored captures, v2 is identical on the eight from 08-12 to 08-25 and would read 2 lines and 1 line on the two 08-29 captures. Gate stays closed at 2 of 6; no row published, and the observations CSV is untouched at 2026-08-25. The first rows captured under v2 will be partly method-attributable and the series method_version column marks them.

• 0.9.4 (2026-08-25) — three same-day captures (20:45, 20:49, 20:57 UTC), run during the crisis-article build after the scheduled morning run did not fire; each preserved in its own evidence directory. Published file = the final capture: 624 reviewed, 5 eligible offers across 4 lines. IronWolf 8TB held $329.99 / $41.25 per TB at two direct retailers at once (Best Buy new this day, Newegg repeat); WD Gold and Exos 7E10 flat. WD Red Pro read $489.11 at Best Buy in capture 1. Docid and seller both match 2026-08-23’s $402.99 row, so that stands as the series’ second clean same-listing move, +21.4% in two days. Capture 3 then read $379.99 at Newegg on the SAME docid, which reconfirms 0.9.3’s finding that a docid can carry different sellers and is not by itself a listing id. Gate stays closed at 4 of 6. Exos SAS/SATA title question still open. Also corrected in this version: the day-one section’s 2024 baseline, which had shipped uncited as “$220-240 through most of 2024” (deep-refinement seg-12 flagged it), now cites camelcamelcamel’s WD8005FFBX history and its $189.99 July-2024 Amazon low; the day-one multiple recomputed 1.7x to 2.1x against the cited anchor.

• 0.9.3 (2026-08-24) — daily capture: 715 listings reviewed, 4 eligible offers across 4 drive lines, the first four-line day in the pilot. Seagate IronWolf 8TB qualified for the first time at $329.99 / $41.25 per TB at Newegg, a new low for the series. WD Gold 8TB flat at $369.99 / $46.25 and Seagate Exos 7E10 8TB flat at $379.99 / $47.50, both on unchanged listings. WD Red Pro 8TB reads $379.99 / $47.50 versus $402.99 the day before, but the seller changed under an unchanged listing id, so it is recorded as not comparable — and that fact also weakens the like-for-like claim made at 0.9.2. The activation gate stays closed at 4 of 6 required lines.

• 0.9.2 (2026-08-23) — daily capture: 734 listings reviewed, 3 eligible offers across the same 3 drive lines. WD Gold 8TB flat at $369.99 / $46.25 per TB at Central Computers and Seagate Exos 7E10 8TB flat at $379.99 / $47.50 at Adorama, both on unchanged listings; WD Red Pro 8TB fell to $402.99 / $50.37 at Best Buy from $457.04 / $57.13, on the same listing id, which makes it the first like-for-like price move this pilot has recorded. The activation gate stays closed at 3 of 6 required lines. Flagged for confirmation, not corrected: the Exos 7E10 row is titled as a SAS drive while the tracked part is the SATA ST8000NM017B.

• 0.9.1 (2026-08-22) — daily capture: 706 listings reviewed, 3 eligible offers across 3 drive lines (WD Gold 8TB $369.99 / $46.25 per TB at Central Computers; Seagate Exos 7E10 8TB $379.99 / $47.50 at Adorama; WD Red Pro 8TB $457.04 / $57.13 at Best Buy). The activation gate stays closed at 3 of 6 required lines and no index row was manufactured. Correction: the answer capsule, the first FAQ, and the latest-capture section still described the 2026-08-12 day-one result while the rendered table already served newer CSV rows — the same drift class the RAM index logged at 0.9.2 and 0.9.4. Prose is now resynced in the same commit as the capture that moves the data.

• 0.9.0 (2026-08-12) — pilot capture and method publication. Two method corrections made against measured evidence on day one, both recorded: price ceiling $400 → $600 (a genuine WD Red Pro 8TB single at Best Buy, $402.99, was rejected by the stale ceiling as its only failing check), and a multi-pack/bundle guard added alongside the raise (measured leak shapes: “WD 2-Pack 8TB Red Plus” $739.98, “Seagate 8TB BarraCuda (OEM Packaging, 2-Pack)” $599.98, “2 x WD Red Plus 8TB” $699.98, 20-pack IronWolf rows — all state the line and capacity, so only the guard keeps them out of the raised band). Run 2’s results were collected through DataForSEO’s tasks-ready endpoint after a queue-delay timeout; no tasks were re-posted and no rows were lost.