HAMR Hard Drives Explained: What Changes at 30TB and Above

Hard drive capacity stalled for most of the last decade. Perpendicular magnetic recording, the technology behind almost every drive shipped since 2006, ran out of room somewhere around 24TB, and the industry spent years extending it with helium fills, more platters and shingled tracks rather than improving the recording itself.

HAMR is the change that ends that. It is the reason 30TB and 40TB drives exist, and it brings a set of buying considerations that did not apply to the drives it replaces.

The problem HAMR solves

Magnetic recording runs into a three-way conflict that has no solution within conventional technology.

To store more data on a platter, the magnetic grains holding each bit have to get smaller. Smaller grains hold their magnetic orientation less firmly, and below a certain size ambient heat alone is enough to flip them, which corrupts data sitting on a drive doing nothing.

The fix is a media that resists flipping — higher coercivity, in the terminology. But a media that resists flipping also resists being written to deliberately, and a drive head can only produce so much magnetic field. Past a point the head physically cannot write to media stable enough to be worth using.

That is the wall. Grains small enough for higher density are not stable; media stable enough at that grain size cannot be written.

How HAMR works

Heat-assisted magnetic recording breaks the conflict by changing the media's properties at the moment of writing.

The drive uses a high-coercivity media, typically an iron-platinum alloy, that is extremely stable at room temperature. A laser in the write head heats a very small spot on the platter for a matter of nanoseconds, which temporarily reduces the coercivity of that spot enough for the head to write it. The spot cools almost immediately and returns to its stable state.

The scale involved is the reason this took two decades to commercialise. The heated area is measured in tens of nanometres, the heating and cooling cycle is over in around a nanosecond, and the laser assembly has to survive billions of cycles inside a sealed drive without degrading. The optical component that focuses the laser to that spot size, the near-field transducer, was the specific engineering obstacle that held the technology back for years.

Why this matters more than the capacity number

The obvious benefit is capacity per drive. The more consequential benefit is capacity per rack unit and per watt.

A drive bay costs roughly the same to power, cool and occupy regardless of what is in it. Moving from 16TB to 32TB in the same bay roughly halves the cost per terabyte of everything around the drive — the chassis, the controller, the rack space, the power draw, the floor space.

For a hyperscale operator that arithmetic drives purchasing decisions outright. For an enterprise running a few arrays it matters less, but it is still the reason capacity jumps at the top end tend to arrive in cloud data centres well before they reach the channel.

Where the technology actually stands

Seagate reached volume first. Its Mozaic platform is HAMR-based, with drives at capacities up to 44TB, and the company reported that Mozaic 4 drives were shipping for revenue to 75 percent of leading global cloud customers by March 2026.

Western Digital has taken a slower route. It has extended energy-assisted PMR to 40TB-class products and has said its own HAMR products arrive in 2027, with a 100TB target for 2029. Toshiba continues to develop HAMR alongside its current nearline range.

Two practical consequences for buyers.

Availability is tight at the top end. Morgan Stanley has projected hard drive supply falling 10 to 15 percent short of demand across 2026, with AI-driven demand growth outpacing supply growth and the shortage persisting into 2028. The highest-capacity drives are allocated to the largest customers first.

The channel lags the datasheet. A capacity announced this year generally reaches ordinary enterprise buyers, and the refurbished market well after that. Availability at a given capacity is a better planning input than what has been announced.

What changes for the buyer

HAMR is invisible to the host. The drive presents the same SAS or SATA interface, the same command set, and the same sector format as the drive it replaces. No driver changes, no controller changes, no firmware requirements on the host side.

What does change is the consequence of the capacity itself, and this is where the planning work sits.

Rebuild windows extend with capacity. A failed drive in a parity array has to be reconstructed from the surviving members, and the time that takes scales with the size of the drive, not with how full it is. A 32TB drive takes roughly twice as long to rebuild as a 16TB one on the same array. During that window the array is degraded and a second failure is unrecoverable on single parity.

This is not a HAMR problem specifically, but HAMR is what makes those capacities common. Our RAID levels guide covers why double parity becomes effectively mandatory above about 8TB per drive, and the argument only gets stronger at 30TB.

Fewer, larger drives means less parallelism. Twelve 8TB drives and three 32TB drives store similar amounts, but the twelve-drive array has four times the spindles serving requests. For sequential archive work that rarely matters. For anything with concurrent access it can matter a great deal.

Failure domains get larger. Losing one drive out of twelve costs you a twelfth of the array's redundancy margin. Losing one out of three costs a third. The blast radius of a single failure grows as drive count falls.

HAMR and SMR are not the same thing

Worth separating clearly, because both are density technologies and both appear on high-capacity drives.

SMR increases density by overlapping tracks, which means writing one track partially overwrites its neighbour, so the drive has to read and rewrite a whole zone to modify data in it. That is why SMR write performance collapses under sustained load and why SMR drives should not go into arrays.

HAMR increases density by making the media itself hold smaller stable grains. Tracks do not overlap, writes do not require zone rewrites, and the performance characteristics are conventional. A HAMR drive behaves like a CMR drive because it is one.

The two are independent. A drive can be HAMR and CMR, which the current enterprise nearline products are. Check the listing rather than assuming either way from the capacity alone.

When a HAMR drive is the wrong purchase

Three situations where the capacity is not the right answer.

The array cannot absorb the rebuild window. If the workload cannot tolerate an extended degraded period and the array is on single parity, smaller drives with a shorter rebuild are the safer configuration until the parity scheme changes.

The workload needs spindles. Concurrent random access across many clients is served better by more drives than by fewer larger ones, whatever the total capacity works out to.

You are expanding an existing array. A parity group uses the smallest member's capacity on every drive, so adding a 32TB drive to a group of 8TB drives gives you 8TB of usable capacity and wastes the rest. Expansion is generally a new group, not a bigger drive in the old one.

Where the workload is sequential and capacity-driven — backup targets, archives, surveillance retention, media libraries — the capacity is exactly what you want, and this is where the technology earns its price.

Sourcing

We supply enterprise nearline drives across capacities in SAS and SATA, in both current and prior generations, with the exact manufacturer part number, the recording technology and the condition stated on every line.

Tell us the array, the controller and what the storage is for, and we will tell you whether the highest capacity available is the right choice or whether a lower capacity in greater numbers serves the workload better. Send a bill of materials through the bulk quote page or email sarah.jane@techsellerusa.com.

Common questions

Tell us the array and the workload and we will tell you which capacity is right, not just which is largest.