Cold Storage at Home: Is LTO Tape Worth It in 2026?
Air-gapped, decades-durable, cheap per terabyte at scale — and saddled with a drive price that decides the whole question for you
Contents
Every couple of years I talk myself into pricing up an LTO tape drive, stare at the number, and put the spreadsheet away again. Tape has a mystique in homelab circles — it is what “real” data centres use, it is genuinely air-gapped, and a cartridge on a shelf will still read in thirty years. Then you look at what a drive costs and the romance meets reality. This is my periodic re-examination, done in early 2026 with current LTO-9 and freshly-launched LTO-10 numbers, of the only question that matters: at what point does tape actually make sense for a home archive, and are you past it?
The short version, which I will spend the rest of the article justifying, is that tape is superb at exactly one thing and the drive price makes that one thing viable only past a scale most homelabs never reach. But the threshold is real, some people are past it, and if you are one of them the case is strong.
What tape is actually good at
Before any economics, be clear about the job. LTO is cold, offline, archival storage: write data, eject the cartridge, put it on a shelf. It is the deepest, coldest tier you can own. It is poor at everything a NAS is good at — random access is measured in seconds of seeking down a physical ribbon, and it wants large sequential writes rather than the constant random access of a live filesystem you poke at all day. So tape competes with cold hard drives on a shelf and with cloud cold storage, and it earns its keep on three properties the alternatives struggle to match all at once.
It is genuinely air-gapped. A cartridge sitting in a drawer is electrically disconnected from everything, which means no ransomware, no accidental rm -rf, no firmware bug, and no remote attacker can touch it. This is a stronger guarantee than a NAS snapshot or even an off-site disk that stays powered and reachable.
It is durable for decades. LTO media is specified for 15 to 30 years of archival life in reasonable conditions, and the format is backward-read-compatible across generations (with a caveat I will get to for LTO-10). A hard drive left unpowered on a shelf for a decade is a gamble; a tape is designed for exactly that.
It is cheap per terabyte at the media level. This is the seductive number, and it is where the analysis has to be careful, because the media is only half the cost.
The 2026 numbers
LTO-9, the mainstream generation as I write this, holds 18 TB native per cartridge (the “45 TB compressed” on the box assumes a 2.5:1 ratio you will never see on already-compressed media like video, photos or encrypted backups — ignore the compressed figure entirely for a media archive). LTO-10 launched in 2025 and pushes native capacity to 30 TB per cartridge. Both are excellent densities.
Here is the catch that decides everything. The cartridges are cheap per terabyte: an LTO-9 tape lands around £110–140 for 18 TB, which is roughly £6–8 per terabyte of media — comfortably under the £12–15 per terabyte you pay for bulk hard drives in 2026. The drive is brutal. An LTO-9 SAS drive is a four-figure purchase, typically £2,500–4,000 new, and LTO-10 drives at launch are scarcer and pricier still. You also need a SAS HBA to connect it, decent cabling, and ideally an internal or external enclosure with proper cooling.
So the fixed cost of entry is a drive priced like a small car’s worth of homelab budget, after which each additional terabyte is cheaper than disk. That shape — high fixed cost, low marginal cost — is the entire economic story, and it means the answer depends completely on how many terabytes you will actually archive.
The break-even, done honestly
Model it simply. Say an LTO-9 drive plus HBA and cabling costs you £3,000 all-in. Tape media costs roughly £7/TB; bulk hard drives cost roughly £13/TB. Every terabyte you archive to tape saves you about £6 versus buying another hard drive for it — but you have to save £3,000 of drive cost first. That is 500 TB of archived data before the tape media savings pay back the drive, if you buy new.
Half a petabyte is far more than most homelabs will ever cold-archive, which is why tape is a hard sell for the median enthusiast. Two things bend the curve, though, and they are why some people still buy in.
First, the second-hand market. LTO-8 and even LTO-9 drives turn up used and ex-enterprise for a fraction of new prices, sometimes £600–1,200, because businesses cycle them on refresh schedules. Buy the drive used and the break-even against disk drops to well under 100 TB, which is a scale a serious media hoarder or photographer genuinely reaches.
Second, and more honestly, the media-cost break-even is not the real reason to buy tape. The air-gap and the thirty-year shelf life are. If you value an offline, ransomware-proof, decades-durable archive of your irreplaceable data, you are buying a property the cost-per-terabyte comparison does not capture, and the drive price is the entry fee for that property rather than a figure you amortise against disk.
How you actually use it: LTFS
Modern LTO is far more pleasant than the tar-and-pray tape of old, thanks to LTFS — the Linear Tape File System — which lets you format a cartridge and mount it as though it were a (very slow, sequential) filesystem, then drag files onto it. Install the LTFS tools for your drive, then format and mount a fresh tape:
| |
If you prefer the classic route, or want maximum sequential throughput for one enormous dataset, raw tar to the character device still works and some people swear by it for backup sets:
| |
Whichever you choose, two operational habits are non-negotiable. Write a checksum manifest (sha256sum) alongside the data so you can verify a restore years later, and store at least a paper note of what is on each cartridge, because a shelf of unlabelled tapes is a shelf of mysteries. Enable the drive’s hardware encryption via stenc if the archive contains anything sensitive, since the cartridge may outlive your memory of what it holds and where it ends up.
There is also a slower creep to plan for: bit-rot and format obsolescence over the decade-plus you expect these tapes to last. Media degrades gently, so a periodic read-back — every couple of years, verified against the manifest — catches a marginal cartridge while you can still rescue it onto a fresh one. And because each LTO drive reads only a generation or two back, a tape written today may need migrating to newer media before the drive that can read it becomes museum hardware. Budget for a re-copy pass every few generations; the media is durable, but the ecosystem that reads it keeps moving on.
The gotchas nobody mentions until you hit them
LTO-9 media initialisation takes hours. LTO-9 introduced “optimised media initialisation”: the first time you write to a brand-new LTO-9 cartridge, the drive calibrates itself to that specific tape, a process that can take up to an hour or two per cartridge and must complete before real use. Plan for it; do not panic when your first write appears to hang.
LTO-10 broke the backward-compatibility tradition. For decades LTO drives read two generations back and wrote one back. LTO-10 reads only LTO-9, with no support for older cartridges at all, and LTO-10 media is its own format. If you have an existing LTO-7/8 archive, an LTO-10 drive will not read it — a genuine consideration if you are choosing which generation to buy into.
The interface is SAS. LTO drives connect over SAS (or Fibre Channel). You need a SAS HBA in the host, which for many homelabbers is an unfamiliar extra card and cable. There are USB-to-tape bridges but they are finicky; a proper HBA is the reliable path.
Shoe-shining kills throughput and tapes. Tape drives want a steady, fast data stream. Feed them too slowly and the drive repeatedly stops, backs up and restarts — “shoe-shining” — which is slow and wears both tape and heads. Stage your archive on fast disk and stream from there, rather than piping directly from a slow network share.
A verify pass is mandatory. Write, then read the whole thing back and check it against your sha256sum manifest before you consider the archive safe and reuse the source space. A tape you wrote but never read back is a hypothesis, the same trap that catches people with disk backups.
Where tape sits in a real strategy
Tape is a cold archive tier, so it complements rather than replaces your warm storage and your regular backups. It is the bottom of the stack under a bulk-media array like the SnapRAID and MergerFS setup I run for day-to-day media, providing the offline, long-term copy that on-line redundancy never can. It is also one very good way to satisfy the off-site, offline leg of the 3-2-1 rule — a box of cartridges at another location is about as air-gapped as an off-site copy gets. For most people, though, the far cheaper way to get that off-site leg is a Raspberry Pi and a disk at a relative’s house, and being honest about that alternative is part of giving tape a fair hearing.
Verdict: is it worth it, and for whom?
For the median homelab in 2026, tape is not worth it, and I want to say that plainly before anyone spends four figures on romance. If you have a few terabytes to protect, cloud cold storage or a rotated off-site disk is dramatically cheaper and simpler, and the LTO drive alone costs more than your entire NAS. The break-even against buying hard drives, at new-drive prices, sits far beyond what most enthusiasts will ever archive.
Tape becomes genuinely compelling in two situations. If you have crossed into serious scale — a photographer or videographer with 100 TB or more of irreplaceable masters, a hoarder with a genuine petabyte ambition — the media economics start working, especially with a used drive, and nothing else gives you that density on a shelf. And if what you are buying is the air-gap and the thirty-year durability rather than pennies per terabyte, tape delivers a guarantee that powered, connected storage cannot, and the drive is the price of admission to that guarantee. Buy the drive second-hand, budget for the HBA and the init time, verify every write, and label everything. If neither of those describes you, keep the spreadsheet closed like I keep doing, and put the money into a second off-site copy of the data you already cannot replace.
Related Content
