Buy what you need now. Relief is not expected before 2027, and waiting mostly risks paying more later with no clear relief date. The current live floor is $10.63/TB on the Toshiba MG Series 8TB Enterprise SATA 8TB.
Daily average price per TB across all tracked SATA enterprise HDDs, Mar 23 to Jul 16. Live data from the DatacenterDisk price tracker — updates every 2 hours. Window reflects the actual history in our database, not a modeled projection.
If the verdict above lands for you, here is the actionable list — the cheapest enterprise hard drives genuinely in stock this moment, ranked by price per TB and updated every 2 hours. Buy links go straight to the live listing; specs link to the full drive page.
| # | $/TB | Drive | Cap | Interface | Cond | Total | |
|---|---|---|---|---|---|---|---|
| 1 | $10.63 | 8TB | SATA-6G | REFURB | $85 | Buy | |
| 2 | $11.33 | 3TB | SAS-6G | New | $34 | Buy | |
| 3 | $11.63 | 3TB | SAS-6G | REFURB | $35 | Buy | |
| 4 | $11.67 | 3TB | SAS-6G | REFURB | $35 | Buy | |
| 5 | $11.67 | 6TB | SAS-6G | REFURB | $70 | Buy | |
| 6 | $12.00 | 3TB | SAS-6G | REFURB | $36 | Buy | |
| 7 | $12.45 | 8TB | SATA-6G | New | $100 | Buy | |
| 8 | $12.50 | 4TB | SAS-12G | REFURB | $50 | Buy | |
| 9 | $13.33 | 6TB | SAS-12G | New | $80 | Buy | |
| 10 | $13.75 | 4TB | SAS-6G | New | $55 | Buy | |
| 11 | $13.75 | 4TB | SAS-6G | REFURB | $55 | Buy | |
| 12 | $14.00 | 4TB | SAS-6G | New | $56 | Buy | |
| 13 | $14.00 | 4TB | SAS-6G | REFURB | $56 | Buy | |
| 14 | $14.00 | 10TB | SAS-12G | New | $140 | Buy | |
| 15 | $14.10 | 4TB | SAS-6G | New | $56 | Buy |
See the full ranking on cheapest per TB, or check where to buy for stock across every capacity and source.
The forecast starts with supply, because in 2026 supply is the whole story. Four data points define the market:
Read the underlying reporting in WD is sold out for 2026, the memory chip shortage, and hard drive prices up 50%.
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 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.
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.
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.
Almost certainly not in a meaningful way. The supply agreements underpinning the shortage are booked into 2027 and 2028, Western Digital is sold out for calendar 2026, and Seagate can meet only 50-66% of near-term demand. Some analysts float a modest correction, but the structural demand from AI infrastructure buildout means most of 2026 is expected to stay elevated. Relief is not expected before 2027.
Buy what you need now. Waiting exposes you to further increases with no clear relief date, because the forces driving prices — hyperscaler demand, forward-booked supply, tariffs, and the NAND crunch — are multi-year, not a passing spike. Prefer refurbished enterprise drives for the best $/TB, and consider LTO tape for cold archive since it is the one medium not spiking.
There is no firm end date, but the earliest realistic relief is 2027, and the parallel DRAM/NAND shortage may run to 2028. The supply that would loosen the market is already contractually committed to hyperscalers through 2028, and new capacity from HAMR drives ramps gradually. Treat 2026 as a year to buy at need, not to wait out.
Because this is a demand shock, not a supply outage. Factories are running near capacity, but seven hyperscalers absorbed essentially all of WD's 2026 nearline output to build AI storage. When demand outstrips even full production and the top buyers lock in forward contracts, retail and channel supply dries up and prices rise — no factory has to close for a severe shortage to occur.
The opposite — SSD prices are rising faster. Since September 2025, tracking shows enterprise SSDs up far more than HDDs because the NAND and DRAM shortage hit flash directly, with manufacturers shifting wafer capacity toward High Bandwidth Memory for AI accelerators. HDDs are up around 50%; some SSD tiers have doubled or more. Neither is falling.
Barely. LTO tape is the one storage medium not spiking in 2026 because its demand base is stable archival use, not AI infrastructure. For cold archive and long-term backup at scale, tape offers the lowest $/TB and is largely insulated from the shortage — though it requires a tape drive, which is economical only above roughly 50TB of archive.
The 2011 Thailand floods took about 28% of global HDD capacity offline and doubled prices within 90 days, but normalized in roughly 18 months because the cause was finite and physical. The 2017-18 DRAM shortage doubled memory prices over 18 months and resolved in about 24 months when new fabs came online. 2026 is different: it is demand-driven and multi-year, so the historical 18-24 month recovery pattern does not cleanly apply.