If you store petabytes, dozens or hundreds of them, your most important decision in 2026 isn’t which drive to buy. It’s whether the software you deploy this year can still buy drives in two years.
That isn’t hyperbole. The hard-drive market has changed shape underneath everyone, and the change rewards one kind of buyer — the buyer who is ready for SMR — with priority allocation, shorter lead times, and better cost per terabyte. Everyone else negotiates for scraps. To see why, stop looking at the market as a buyer and look at it the way the three companies who make the drives do.
Look through the HDD makers’ eyes
The vast majority of bytes in large-scale storage still land on spinning disk, and nothing at flash pricing changes that this decade. So, the terms are set by the three remaining HDD makers: WD, Seagate, and Toshiba.
Through the COVID cycle, hyperscalers ordered aggressively ahead of demand, then stopped – leaving all three vendors with production they couldn’t wind down without real financial pain. That lesson stuck. When the AI wave hit and nearline demand went vertical (training sets that regenerate daily, petabyte checkpoints, inference output nobody deletes), the vendors refused to chase it with speculative production.
The shift nobody put in a press release: build-to-order
The HDD industry has fundamentally changed its playbook. Gone are the days of building against optimistic forecasts. Today, manufacturers prioritize firm, effectively non-cancellable orders, committing production capacity only where demand is guaranteed – and holding back investment in new manufacturing lines and R&D until those commitments are in place.
The signals are unambiguous. In February 2026, WD CEO Irving Tan said the company was “pretty much sold out for calendar ‘26,” with firm purchase orders across its top seven customers and long-term agreements into 2027 and 2028. Seagate’s Dave Mosley said nearline capacity was “fully allocated through calendar year 2026.”
The consequence: committed capacity goes to the largest accounts first. Everyone else – enterprise, sovereign-cloud, public-sector, and data-intensive buyers who aren’t a top-seven hyperscaler – is served on what’s left. That’s the shortage, and it’s the operating model now, not a blip.
Can’t make more drives? Make bigger ones
HDD manufacturing is constrained by physical production capacity. A factory can produce only a relatively fixed number of drives, while expanding manufacturing requires significant capital investment and years of planning. As a result, vendors seeking growth aren’t focused on shipping more drives—they’re focused on shipping more terabytes per drive, increasing areal density and pushing storage capacities to new levels.
The cleanest way is SMR. Shingled Magnetic Recording overlaps tracks like roof shingles, yielding roughly 20–25% more capacity per drive at essentially the same cost of goods as a conventional (CMR) drive. Every CMR line flipped to SMR grows sellable capacity — no new factory, no new units. Seagate put exactly this on screen at MSST 2026:
Seagate, MSST 2026 (“Seagate: HAMR/SMR,” R. Mendonsa). The green SMR line pulls away from the blue CMR line — the capacity uplift widens back toward ~20% as HAMR ramps.
The chart tracks CMR against SMR capacity year over year; what matters is the gap — the SMR uplift. It compressed early in the HAMR transition and is now widening back toward ~20%. And it isn’t one vendor. WD pioneered cost-managed UltraSMR and is qualifying 40 TB UltraSMR for volume production in H2 2026. Toshiba began sampling 30–34 TB SMR nearline drives in April 2026 — SMR ahead of its CMR equivalents. All three treat SMR as a bestseller and want it beyond the hyperscalers.
Vendors we talk to put SMR at roughly 60% of the nearline mix today, on a path toward ~80% within about two years. Within a couple of years, roughly 8 of every 10 nearline drives will be SMR.
In a shortage market, SMR is your priority pass
Here the manufacturer’s incentive and the buyer’s line up. When drives are scarce, a large buyer cares about three things in order: allocation (can I get them at all?), lead time (when do they ship?), and price (usually the price at shipment, not today’s quote — so a rising market means you pay tomorrow’s number). All three are painful right now.
What relieves all three at once? Choosing the drive the vendor most wants to build. Commit to SMR and you help the maker convert fixed drive-count into extra sellable capacity that would otherwise be lost — so you move to the front of the queue.
This isn’t just our reading — it’s how the vendors now run the business. They keep score in capacity, not units. Both WD and Seagate have stopped reporting HDD unit shipments altogether: across full-year 2025, units rose ~0.4% while exabytes shipped grew ~22% (Coughlin, Forbes). When the scoreboard is exabytes, the buyer who helps book more exabytes per drive goes to the front of the line. This is how Leil frames the supply picture:
The capacity squeeze puts SMR first in line. Demand is outrunning supply and vendors can’t just make more drives. Flipping CMR to SMR is how they grow sellable terabytes — so SMR is the tier they most want to move. Vendors prioritize SMR, and an SMR-native buyer gets better allocation, shorter lead times, and lower $/TB.
The economics reinforce it: the buyer wins on cost per usable terabyte, and the vendor books margin without building anything. In a market this tight, the outcome collapses toward a binary: you are either an SMR buyer with availability, or you are waiting.
The catch – it’s a software decision, not a hardware one
Benchmark your workloads in our remote lab, scope a proof-of-concept, or just talk architecture with someone who has been inside the drive firmware:If SMR is the winning hand, why isn’t everyone playing it? Because SMR drives don’t behave like CMR drives. They are host-managed and sequential-write; they need software that speaks the drive’s native physics – zone awareness, sequential data flow, workload shaping. Hyperscalers built that years ago. Most enterprise stacks didn’t: they were architected around decades-old random-access logic and treat a 32 TB SMR drive like a 4 TB CMR drive from a decade ago. Bolt SMR onto a CMR-only architecture and you don’t get production-grade support – you get a science project.
This is the challenge. The software you standardize on this year isn’t a one-year decision – a serious deployment runs five years and will need capacity expansion. Lock into a CMR-only stack now and here’s where you land in 24 months: SMR is ~80% of supply, you can natively use the other ~20%, and expansion runs straight into extra cost and availability problems.
The way out isn’t to abandon CMR – it’s to refuse the architectural corner. Choose a stack that is drive-type-agnostic and vendor-agnostic: CMR and SMR side by side, in one cluster, across all three makers, absorbing the roadmap without a forklift upgrade. Then, when SMR is the better bet on supply and price – which is now – you simply take it.
The strategic call
Most of today’s storage headlines are dominated by ”memflation”—the sharp rise in DRAM and flash prices—casting hard drives as the low-cost alternative. But the bigger story is unfolding elsewhere. The industry’s HDD supply landscape is changing, and even fewer conversations address the software layer that determines whether organizations can fully capitalize on the latest high-capacity drives. That’s where the real opportunity lies. For data-intensive organizations, the path forward is becoming increasingly clear:
Assume the shortage is structural. Plan procurement as if committed, build-to-order allocation is permanent – because it is.
Get SMR-ready now. Run a proof of concept this quarter. Being a credible SMR buyer earns priority allocation, shorter lead times, and better $/TB while others wait.
Choose software that keeps every door open – CMR and SMR, all three vendors, mixed generations in one cluster – so your 2026 decision doesn’t cap you out of 80% of the market in 2028.
The next few years of storage belong to the organizations that read the market the way the drive makers do – and deploy the software that lets them act on it.
Author: David Gerstein David Gerstein is the CTO of Leil, an Estonia-based storage infrastructure company building HDD-native software for the exabyte era. Leil is in active co-engineering partnership with Western Digital, validated across the WD, Seagate, and Toshiba drive ecosystems.
