back to article The trendline doesn’t look good for hard disk drives

In early May, independent digital storage analyst Thomas Coughlin shared news of falling sales and revenue in the first quarter of 2025, continuing a trend that started in around 2010. Coughlin cites data from that year showing around 600 million annual hard disk shipments. In 2025 he thinks around 150 million units will make …

  1. billdehaan

    There is a lot of "non-active" data out there

    There was a joke decades comparing the speeds of different dedicated music processors, where the market leader bragged that "Our processor can play the Minute Waltz in 466ms; our nearest competitor takes more than 800ms".

    Of course, in the real world, the Minute Waltz should play at a speed of... one minute (hence the name). The value to the customer of the improved performance was somewhat questionable.

    SSDs are fantastic for performance, they're also shock proof (a major benefit in industrial areas where vibration kills hard drives like no one's business), and both the form factor and power consumption benefit from not requiring a motor.

    But, for all the benefits, SSDs have some downsides, too. Cost, for one. While the small (1TB) size SSDs are price competitive with HDDs, at higher densities, HDDs are significantly cheaper. At 4TB, the HDD is 1/4th to 1/3rd the price of an HDD, and the higher you go, the cheaper HDDs are relative to SSDs.

    The other issue is that SDDs aren't great for long term storage. A damaged HDD can often be recovered, even if only partially, but I don't think I've ever seen a damaged SSD recovered. Once they go, they're gone.

    I absolutely configure all desktop PCs with SSDs, but my video server has 20TB of videos on HDD, not SSD. And backup solutions aren't SSD based, either.

    With more desktops being replaced with laptops, the trend line for HDDs is definitely going down, but until SSD reliability, MTBF, recoverability, and cost become competitive with HDDs, I don't see data centers phasing hard drives out any time soon.

    1. find users who cut cat tail

      Re: There is a lot of "non-active" data out there

      Exactly. You have SSDs for everyday fast access. But your longer-term high-capacity storage is unlikely to be SSDs. It does not have the capacity/price ratio. And trust.

      We all have HDDs from the era before SSDs became common still happily spinning and storing data (even though the capacity is usually laughable now). There are practically no really old SSDs still working. The longevity might be improving but gaining trust in storage takes decades.

      1. Lon24 Silver badge

        Re: There is a lot of "non-active" data out there

        Indeed - I clone or rsync all my SSDs to ancient similar capacity HDDs. Too slow for active use but, basically, free backup as they are fully written down. But buying new? No chance.

      2. Woodnag

        PC storage

        For Windows, I prefer a 128GB SSD as the C: drive, an HDD for D:, and locate the user standard directories (Docs, Video, Pictures, Temp etc) on D. Easy to backup user stuff by just copying D:.

        If the SSD isn't encrypted, the unused areas compress well when imaged, so "sudo dd bs=4M if=/dev/sda | sudo gzip -c > Date_C_drive.gz && sync" from a live Linux USB provides a backup similar in size to the used storage, and 128GB doesn't take long. If encrypted, don't bother with gzip. And the backup can go on the D drive initially, so fast.

        1. Anonymous Coward
          Anonymous Coward

          Re: PC storage

          For Windows, I prefer format c: followed by booting from a Linux distribution

      3. Anonymous Coward
        Anonymous Coward

        Re: There is a lot of "non-active" data out there

        I don't actually have any pre-SSD era HDDs in use. I went SSD for system drives in about 2009 and the first drive still works. Mechanical drives from that era don't have enough capacity to be worthwhile to me. Sure, I have couple of 1TB external drives, but for actual storage I have no HDDs below 4TB.

    2. MachDiamond Silver badge

      Re: There is a lot of "non-active" data out there

      "Once they go, they're gone."

      NAND chips will often fail shorted which bricks the whole unit and the only way to try and recover something is to remove chips (if more than one) to find the shorted one and hope the others have complete files. It's a big issue with Apple's approach to having proprietary storage that can often brick the whole computer when it fails with no easy way (aside from sending to Apple) to repair.

    3. Anonymous Coward
      Anonymous Coward

      Re: There is a lot of "non-active" data out there

      Somewhat agree. But I dont think HDDs are nearly cheap enough for their relative performance.

      Price per GB doesnt matter anymore. I mean if you're nickel and diming sure, but in the grand scheme of things its a meaningless comparison.

      For me its about how long it takes to move terabytes rather than how many I can store.

      The only way I'd consider spinning rust over SSD at scale is if it was outlandishly cheap. Say an 20TB drive cost about £25-50. I would scale that up for performance in a heartbeat. Because I could buy enough of them for the performance to be worthwhile. However the reality is a drive that size costs 10x that at least. Which means I can't buy as many and big block storage on hard drives becomes pointless because it is too expensive to reach a meaningful level of performance.

      1. mirachu Bronze badge

        Re: There is a lot of "non-active" data out there

        Show me 16TB SSDs that are sensibly priced for consumers. I have 32TB on my desktop, 4 of that SSDs, a few hundred gigs of that L2 cache for the spinning rust. 32TB of SSD is prohibitively expensive and makes no sense anyway.

        Point is, you can take the worst edge of HDD worst case scenarios away with caching, as long as your use case is at least somewhat sensible (completely random small accesses aren't generally sensible and if you have tons of those you ultimately might want memory, not disk). For sequential access, I figure the 250MB/s+ modern HDDs get is enough for personal use.

    4. Malcolm 5

      Re: There is a lot of "non-active" data out there

      I know i am being fussy

      "Of course, in the real world, the Minute Waltz should play at a speed of... one minute (hence the name)"

      The minute waltz is called that because it is small.

      Wikipedia look like a good summary of my concern

      "The piece is given the tempo marking Molto vivace. Although it has long been known as the "Minute" Waltz, its nickname was intended to mean "small" in the sense of a "miniature" waltz, given by its publisher.[3] Chopin did not intend for this waltz to be played in one minute. A typical performance of the work will last between 1+1⁄2 and 2+1⁄2 minutes"

      1. Martin an gof Silver badge

        Re: There is a lot of "non-active" data out there

        Of course, Radio 4 perpetuates the myth, using the Minute Waltz as the theme music for Just a Minute.

        M.

      2. headrush

        Re: There is a lot of "non-active" data out there

        It helps if the word is pronounced correctly.

        The "my newt" waltz does not describe an amount of time.

    5. MachDiamond Silver badge

      Re: There is a lot of "non-active" data out there

      "There was a joke decades comparing the speeds of different dedicated music processors, where the market leader bragged that "Our processor can play the Minute Waltz in 466ms; our nearest competitor takes more than 800ms"."

      There's also the issue that a massive change in one spec can be completely meaningless in the real world. It's like 0-60mph stats for a car where better is the enemy of good enough. Not just in that it doesn't matter, it can be too much performance that can get one in trouble.

      SSD's have their strengths. I use one in my MacPro as the system driver and another as my working disk for photos and video. Once a project is completed, those files get sent over to some regular HDD's. The cost of big SSD's is far too much to be financially efficient. I change over my HDD's every couple of years with the old ones becoming archive drives where I backup the most recent files and go back as far as there is space. I wind up with things being backed up over multiple drives as I'll prune away the obvious dead files to have plenty of space. The files left on the computer right now tend to go back too far. I have 5 years of photo jobs in Lightroom when I really only need to go back 2 at the most. Software goes back even more since it takes up far less space and I don't think I've deleted any of the project development jobs as they are handy to rob from if they are to hand.

    6. Annihilator Silver badge

      Re: There is a lot of "non-active" data out there

      Yep, over the years I've had many many MANY spinning rust start to throw out errors or weird noises that make me replace it - usually successfully.

      Every single SSD I've ever owned has imploded without warning and utterly catastrophically.

  2. AceRimmer1980
    Thumb Up

    Until SSD technology can exceed HDD's for robustness and longevity, it'll be spinning rust for me in my file server.

    1. hoola Silver badge

      SSDs do in physical terms. It is in data and electronic terms they don't. This is mainly down to the way they fail.

      1. werdsmith Silver badge

        A failed drive is a failed drive, and whether it is spinning platter or SSD. In storage systems we have redundant units and hot swap. We’ve always done it this way since long before SSD became popular. If you find yourself needing the services of a data recovery company it is because of poor practice.

        If your computer is bricked because it uses soldered in storage, that is because if the way it is constructed, not the storage tech. OK so the storage tech enables the soldering in, but if it’s important data then back it up regardless of primary storage.

      2. Anonymous Coward
        Anonymous Coward

        > SSDs do in physical terms. It is in data and electronic terms they don't. This is mainly down to the way they fail.

        Failure isn't the only issue. SSDs lose data over time when unpowered (the average data retention period of SSDs is between 6 month to 3 years), which makes the poor choices for long-term storage such as data archival.

  3. nobody who matters Silver badge

    Hmmmmmm........Gartner..........again!

  4. Hurn

    PCIe connectivity should be the lede, not NVMe

    "NVMe (nonvolatile memory express) is the protocol that connects SSDs directly to a server's CPU over the Peripheral Component Interconnect Express (PCIe) bus, greatly speeding data transfer speeds. If hard disks can use NVMe, they'll be relevant in more roles."

    While NVMe has less overhead than SCSI, it's the channel that sets the speed limit.

    Getting HDDs OFF OF SATA III (6Gbps) is the key to allowing them to be faster.

    SAS 12G helps. as does SAS "dual channel" (which, should work great with dual actuator / head array HDDs)

    But, it's getting HDDs onto PCIe x4 version 3, 4, maybe even 5 that's really going to raise the external speed limit.

    (and, mix & match width vs version: PCIe x1 version 5 vs PCIe x4 version 3)

    Internal speed limit comes down to the same factors it ever has:

    Disc speed (RPM)

    Data density (bits per linear unit of measure)

    Number of heads performing I/0 at once (this metric is where tape rules!)

    Domain latency/hysteresis (how long it takes the head to read the bit - interdependent with disk speed, density, and tricks)

    Smart caching (YMMV according to workload)

    1. brainwrong

      Re: PCIe connectivity should be the lede, not NVMe

      "Getting HDDs OFF OF SATA III (6Gbps) is the key to allowing them to be faster."

      HDD's are still limited by how fast the bits go under the read/write heads. Only 1 head is used to read or write at a time, which I assume is down to tracking tolerances.

      Modern HDD's are nano-precision marvels! I have some 7200rpm drives that max out at around 240MB/s on the outer tracks, that's 2MB on a track about 250mm long, so 8 bytes/micron, 64 bits/micron, so bits are about 16nm long. Inner tracks are half the length, so a mean of 1.5MB per track. 1TB per platter side suggests around 670,000 tracks, spanning a sweep of about 20mm, or 33.5 tracks/micron, or about 30nm apart. It's working at similar scales to transistors on modern chips (which aren't as small as 2 or 3 or 5 or 7 nm that process names might suggest), which require some rather expensive equipment to make.

  5. Michael Hoffmann Silver badge

    Meh...

    ... my home kitchen sink (*) server is still built around spinning rust and will likely remain so. Every so often I upgrade with the hand-me-down mobo, RAM and CPU when I build a new desktop. The RAID controller and disks simply get plugged into the new system and off we go. It's cheap and does the job - no cloud involved! Also every so often a disk fails in the array, order a new one for tuppence, swap, sync, done.

    (*) ESX with VMs running Samba, DNS, Ark Survival (no, really!) and backups

    1. Lee D Silver badge

      Re: Meh...

      I'm about to buy a NAS for myself, and others for work, and to be honest, I'm looking at ones that include several M.2 NVMe slots.

      Because, for a start, I already have a bunch of them lying about that are too small for individual usage, but would make a great cache for a NAS to stop the disk spinning more than they need to.

      But also, you can just get NVMe NAS nowadays, and the NAS I'm looking at can make RAIDs out of multiple NVMe's too.

      It's all going that way. Every computer in work uses a M.2 stick. The main servers are M.2 for boot drives and SSD for caching. My laptop is 2 x M.2. My RPi's all have NVMe boot support with a tiny cheap adaptor for hardware.

      And just looking at the size of things, you can put 4 M.2's in the bottom of a NAS and add zero size to the device at all. You can get a slimline router-size thing that's an entire RAID NAS for NVMe.

      Disks are dying, and another few years they'll be dead. I only buy hard drives for large RAIDs nowadays, and those are all backed by SSD caching and fast approaching the point where the SSDs are nearly as cheap.

      So I'm looking at a 10-bay NAS with 4 extra slots for NVMe. That should be enough to throw all my old stuff in, including old spare disks, SSDs and NVMe's, and then if I need to I can keep using that for a long time with a handful of cheap adaptors and bring the modern tech into the old disk bays (just not at NVMe speed, but who cares?).

      As far as I'm concerned, disks are on the way out and even this is just a nod to using the old stuff I have lying around, which is starting to become as much NVMe SSD as if it hard disks. When I upgraded my laptop to WD Black M.2's, the original sticks... they're just lying on a shelf. Why not make use of them? Cache at first but, you know what? Storage just as much when I later replace those WD Blacks.

  6. The Dogs Meevonks Silver badge

    So make large capacity SSD's cheaper then

    I'd happily swap out my mediaserver drives (approx 68TB) for SSD's... if they weren't so stupidly expensive.

    Right now, I can get a 16TB HDD for about £195, you cannot buy consumer SSD's in that capacity, with the largest being 8TB and costing well over £500, and if you want an 8TB M.2 NVME drive, you can pretty much double that price again. Meanwhile, my HDD's are used for storage only, they get written to and rarely over written, so they last a very long time... I have a couple of 6TB drives with 80-90000 hrs of uptime on them. and the largest 14TB ones I currently have are over 30000hrs now. In fact, I tend to buy seagate external drives and rip the HDD from them because it's so much cheaper to do so. My biggest issue is that my mediaserver has no raid or backup for the media on it... and I'd love to be able to afford to buy 6 or 7 20TB HDD's and do a software raid with parity for peace of mind, because my BDrom died recently and I'm not going to be able to rip my own movies from my own discs anymore unless I can find a 13yr old brand new pioneer drive without the firmware that now restricts this ability.

    I've currently got 7 HDD's in my mediaserver, in a mix of capacities. If I replaced them all with 8TB SSD's it would cost me over £2500 and I'd lose about 18% of total capacity.

    As I've been saying for nearly a decade now... until price per GB falls to sensible levels for SSD's, HDD's aren't going anywhere... SSD's are cheaper to produce, have no moving parts... larger capacity SSD storage isn't a problem... it's that they're constantly pushing 'speed' over capacity when that's not what people want or even need. There's this delusional thinking that no one needs large capacity storage because no one is allowed to actually own anything any more... you can only access it in the cloud and stream it.

    1. ridley

      Re: So make large capacity SSD's cheaper then

      UNRAID can help you have some protection for your data.

      It's saved mine in several occasions.

      (144Tb R720 and a disk shelf)

  7. blu3b3rry

    Most of the spinning rust I keep around now is simply as backups. Nice big 4TB HDD keeping snapshots of everything on my desktop. The other tower PC under the desk also uses NVME for its Linux Mint install, alongside an old WD Raptor HDD with a W11 install for the rare occasions it's needed. I was rather impressed with its performance booting W11 although you definitely know when its running at 10,000 rpm!

    Even my ancient 2008 HP laptop has a 1TB cheapo SSD thrown in it, I think it cost me £20 on an Amazon sale. It's definitely not quick as SSD's go but given the laptop only has a SATA II interface that doesn't really matter.

    ....that said, I do miss the gentle click and whine of an HDD letting you know how hard the computer is working.

  8. StrangerHereMyself Silver badge

    Long term storage

    I still use HDD's because I don't trust SSDs for my long-term storage i.e. put it in a box somewhere and put away for years and years. After a couple of years your SSD data is gone, every last bit of it. A HDD will still be readable after decades and most if not all of the data will be recoverable.

    I suspect we'll see many heart-broken families in the future who'll have lost all their precious family photos and videos due to their SSDs failing.

  9. Kevin McMurtrie Silver badge

    Challenger gone

    Synology is deep in their 3rd year of enshitification. There no reason to use them instead of off-the-shelf hardware with standard software.

  10. Missing Semicolon Silver badge

    What about write limits

    HDDs can be written to all day, every day (as swap, for example). SSDs cannot tolerate long-term write-heavy loads.

  11. steviebuk Silver badge

    Bullshit

    All the people on datahoarders on reddit talk regular about getting lots of drives and large ones for data hoarding needs. Even religiously replacing them after 5 years, even if all tests show they are working fine.

    I'd be doing the same, if I could afford them. Biggest I have are 2x10TB but they are both refurbs.

    Spinny drives aren't going anywhere soon.

  12. David Newall

    Evaporating data

    As I discovered when I wondered what was on an old memory stick, the data needs to be periodically refreshed. You can't just leave them in a drawer and expect to be able to read them in 10 years, or however long "archive" is for.

  13. jonfr400

    Because of new computer cases

    All new computer cases have that problem there is no space for hard drives. Unless its a ssd or m2 hard drives. There is sometimes space for one or two hard drives at most. In many cases there is no space for normal hard drives. Making it impossible to have many hard drives installed.

    They clearly want to move us all to the internet cloud.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like