Nobody likes talking about drive failure. The industry measures it because the industry has to — data centres replace drives on a schedule, and the schedule costs money — but the consumer version of the conversation usually stops at a shrug. Eventually they all fail. True. Also uselessly abstract.
Here is a better version. There are two shapes of drive death that matter for a Mac owner. The mechanical kind, which follows a well-known curve that we can name and show you. And the solid-state kind, which is not about moving parts at all but about how many times you've written to the thing. Both are knowable. Neither is inevitable in the way folklore makes out.
This is written by the team that builds a backup product, so yes, the conclusion will sound familiar. But the data here is public, the sources are cited, and you can go and check the numbers yourself. If you come away thinking you don't need a backup, we've still done our job — you've thought about it.
Chapter 1 — The bathtub curve
The first thing a reliability engineer will tell you is that mechanical failures don't happen at random through the life of a drive. They cluster. There's a bump at the very start, a flat middle, and a rising tail. Plotted over time it looks like the cross-section of a bathtub, and that's the name the industry uses.
The bump at the start is infant mortality. A handful of drives in every shipment arrive with latent manufacturing defects — a misaligned head, a bad platter batch, a firmware quirk — and those drives tend to fail within the first year. The flat middle is the quiet stretch where nothing much happens unless you're unlucky. The rising tail is wear-out: bearings wearing, lubricants drying, platter surfaces degrading. After roughly year four, drives start dying because they are old, and the curve climbs from there.
Hard drive failure rate by age (illustrative)
The classic bathtub shape, compressed: infant mortality inside year 1, a low and flat middle, a steepening climb from year 4 onwards.
What's striking about this curve is not that it rises but how it rises. In fleet data published by Backblaze — more than a decade of quarterly reports across hundreds of thousands of drives — the jump between year four and year seven is not incremental. It is roughly 3x on a per-year basis, and the oldest tranches in their fleet have crossed 9% annualized failure rates on some models. That's one in eleven, every year, for a drive that's passed its seventh birthday.
If your external 5-TB USB drive is the one you bought for the wedding shoot in 2019, it is now sitting on the right-hand side of this curve. It is not going to fail this afternoon. But the probability it fails in the next twelve months is materially higher than it was two years ago, and the probability keeps climbing the longer you leave it.
Chapter 2 — SSDs, by the numbers
Solid-state drives don't have bearings to wear out, so some of the bathtub logic doesn't apply. What they have instead is a finite number of write/erase cycles per memory cell, and a well-characterised way of degrading as those cycles accumulate. The spec sheet for every consumer SSD lists this as TBW, Terabytes Written — how much data you can write to the drive before the manufacturer withdraws the warranty.
Typical consumer tiers rate entry-level 256 GB drives at around 150 TBW, mid-range 1 TB drives at 600 TBW, and higher-endurance 2 TB drives in the region of 1,200 TBW. Enterprise SSDs are rated in petabytes, but those aren't what's in your Mac. What's in your Mac is the consumer part, and the consumer number is the one to keep in mind.
Consumer SSD endurance: rated vs. observed (illustrative)
Manufacturer TBW ratings are conservative. Long-running field studies have consistently shown consumer SSDs outlasting their rated endurance by a factor of 2-3x before first-failure errors appear.
The good news is that rated TBW is conservative. Fleet studies going back to 2015 have shown consumer SSDs routinely outlasting their rated endurance by two or three times before the first uncorrectable errors appear. A 1 TB drive rated at 600 TBW is usually still working at 1.2 PB.
The less-good news is that the drive in a modern MacBook Pro is not replaceable. Since the 2016 Retina redesign, Apple has soldered flash directly onto the logic board on most Mac laptops. When that storage wears out, you don't swap the drive — you swap the machine, or you pay Apple for a logic-board replacement that costs more than most people want to spend on a four-year-old laptop.
There is a way to see how close your own Mac's SSD is to the end of its rated life: the drive reports its wear level through SMART, and a tool like DriveDx will translate that into a percentage. On a healthy Mac that gets used for browsing, email, and the occasional photo import, SMART wear climbs so slowly you'll replace the machine before you notice. On a Mac that's been a video scratch disk for six years, it climbs faster.
One caveat on SMART: it is not a crystal ball. The same fleet studies that documented consumer-SSD endurance also documented that some drives fail without any SMART warning at all. The 2016 Google/Toronto paper found that roughly a third of drive failures gave no advance indicator — the drive reported full health right up until the moment it returned uncorrectable errors. SMART is useful as a trailing indicator of wear, not as a leading indicator of failure. A clean SMART report is a reason to keep using the drive. It is not a reason to skip backing it up.
Chapter 3 — What kills Macs
The fleet-level data we've just walked through is about drives in racks — temperature-controlled, vibration-free, running 24/7 on clean power. A Mac is not in a rack. It's in a bag on a train. It's on a kitchen counter next to a glass of water. It's in a studio with a cat. The failure modes are different, and the data we have on them is thinner.
Our own 2026 Mac Data Loss Report — a survey of Mac users across the creator, professional and small-team cohorts — points to a breakdown that doesn't look much like a server room.
What causes data loss on a Mac, share of incidentsPlaceholder data
External drive failures outrank internal drive failures by a noticeable margin, and physical damage from laptop drops / liquid is roughly a fifth of all incidents.
The headline is that external drive failure is the single biggest category. Bus-powered USB SSDs and consumer portable spinning drives take a physical beating that the Samsung-rated TBW tables don't anticipate. They get dropped. They get sat on. They get unplugged while busy. Their connectors develop intermittent faults. The drive mechanism inside might still be fine, but the housing or the cable or the port can fail independently.
The second headline is that the laptop form factor has its own failure physics. A MacBook Pro on a desk is, statistically, in about the same place as a server. A MacBook Pro travelling between locations is in a materially different risk pool. Drops cause logic-board damage that knocks out the soldered SSD. Liquid ingress does the same. Thermal faults from charging in a closed bag cause intermittent boot failures. The drive isn't dead; the machine can't reach it. From a recovery point of view the distinction is academic.
Ransomware is a smaller slice of the Mac pie than it is on Windows, but it is not zero. The LockBit 3.0 leaks in mid-2023 included a functional macOS build, and commodity ransomware targeting macOS has been live on criminal marketplaces ever since. For creative professionals whose working set is many terabytes of irreplaceable source files, a single encrypted home folder is a catastrophic event.
User deletion is the slice most people underweight until it happens to them. Not the accidental drag-to-trash: that's usually recoverable until the Trash empties. The more painful version is the confident Command-Delete on the wrong folder, the "clean up downloads" session that takes a year of invoices with it, the sync client that propagates a local deletion to every device you own before anyone notices. Human error is a database-grade problem — you need snapshots, not mirrors, because the whole point is being able to go back to a point before the deletion.
Theft and loss round out the picture. Airport lounges, trains, rental cars, and coffee shops account for more Macs than most people think. FileVault prevents the thief from reading your files; it does nothing to return them to you. If the only copy of your work was on the laptop in the bag that got stolen, the encryption is a cold comfort.
Chapter 4 — How long do you have?
With all of the above in hand, the useful question isn't will it fail — yes — but when should I plan for it. A decision aid we'd actually send to a friend:
None of these rules are absolute. A cheap drive can die in its first month; a well-made drive can pass year eight without a complaint. What the rules are good for is planning — what do you replace before it fails, and what do you back up so that when something does fail, the failure is annoying instead of catastrophic.
Chapter 5 — What to do today
Three concrete things, in order of how much they reduce real-world risk.
- Identify your single points of failure. Walk through every drive plugged into your Mac and every folder on the internal disk. For each one, answer: if this died right now, how much work would I lose? If the answer is anything more than a shrug, that drive needs a backup somewhere else. Most people discover one or two of these within five minutes.
- Test a restore. If you already have a backup — Time Machine, a cloud provider, an external clone — restore something specific from it. Not a sample file from the demo folder: a real working file you care about. If the restore works, you have a backup. If it doesn't, you have a habit of making backups, which is not the same thing.
- Start backing up. If you've been putting this off, the friction is lower than it has been at any point in the last decade. A cloud-backed Mac with end-to-end encryption, continuous versioning, and a tested restore path is a one-evening setup. Start a 14-day macup trial, point it at everything that matters, and let it run. Then come back next week and test a restore.
There's a quieter fourth thing, which is: the bathtub curve is not a reason to panic. It is a reason to pay attention. The drives in the fleet studies were fine until they weren't. Occasional vigilance is the whole job.
Sources
-
Backblaze, Inc. Drive Stats. Quarterly public reports, 2013–2024.
https://www.backblaze.com/cloud-storage/resources/drive-stats - Pinheiro, E., Weber, W. D., Barroso, L. A. (2007). Failure Trends in a Large Disk Drive Population. USENIX Conference on File and Storage Technologies (FAST '07).
- Meza, J., Wu, Q., Kumar, S., Mutlu, O. (2015). A Large-Scale Study of Flash Memory Failures in the Field. ACM SIGMETRICS Performance Evaluation Review, Vol. 43.
- Schroeder, B., Lagisetty, R., Merchant, A. (2016). Flash Reliability in Production: The Expected and the Unexpected. USENIX FAST '16.
- Apple Inc. MacBook Pro (2016–present) service and repair documentation, on soldered flash architecture and logic-board-level storage replacement.
- macup. 2026 Mac Data Loss Report (placeholder edition). /backup/report/mac-data-loss