Cold storage and archive tiers for finished projects in 2026

Finished projects should not live on your fast, expensive editing storage. The moment a job is delivered and approved, its footage stops earning its keep on the NVMe and the working NAS, and it becomes archive: something you almost never touch but cannot afford to lose. The right move is to push it down to the cheapest medium that still meets your retrieval needs, and the three real choices in 2026 are LTO tape, cold object storage in the cloud, and a shelf of cheap spinning disks. Each one trades money against how long you wait to get a project back. This is how to pick, and how to build a tiering plan that does not surprise you on the day a client calls back two years later.

Why finished projects need their own tier #

Active editing storage is priced for speed, not for sitting still. Enterprise CMR hard drives that were a commodity 18 months ago have roughly doubled: a WD Ultrastar 18 TB that sold around $180 in early 2024 was pushing $400 by mid-2026, and a 12 TB Seagate Exos went from about $105 to $250-plus over the same window (HDDHunt, checked Jun 2026). The cause is not a mystery. AI buildout has hyperscalers buying enterprise drives in bulk, and WD and Seagate are steering supply toward high-margin datacenter contracts. Keeping a delivered, never-touched project on that storage is like paying for a parking spot in a downtown garage for a car you drive once a year.

Archive storage solves a different problem than your working tier. It optimizes for cost per terabyte and durability over years, and it accepts that getting the data back will take minutes, hours, or a day. The discipline is simple: define when a project is "done," and on that day move it off the working set. Everything below is about where it should land. If you are still deciding what the working tier itself should be, the HDD vs SSD vs hybrid cost math and the storage appliance roundup cover that; this guide picks up where a project leaves the active pool.

Option one: LTO tape #

Tape is the classic archive medium, and 2026 is a real moment for it. The LTO Program finalized LTO-10 with a 40 TB native cartridge specification in November 2025, with the 40 TB media scheduled to ship in Q1 2026 alongside the 30 TB cartridges already on the market (StorageNewsletter and lto.org, checked Jun 2026). A single cartridge holds 30 or 40 TB native, and the cartridges themselves are cheap: $275 to $315 each as of mid-2026 (TechRadar, checked Jun 2026). At 40 TB native, that is roughly $7 to $8 per terabyte of media, and the tape carries an up-to-30-year archival life if you store it at the recommended 16-25 C and 20-50% humidity (Quantum LTO media specs, checked Jun 2026).

The catch is the drive. A standalone MagStor LTO-10 full-height SAS desktop drive lists at $13,299 (MagStor, checked Jun 2026). That is the entry fee before you have written a single byte. Tape only makes financial sense once your archive is large enough to amortize that hardware: at sub-$8/TB media, a five-figure drive pays back against cloud only across many tens of terabytes archived and held for years. Tape also gives you something cloud cannot: a true air gap. Once a cartridge is ejected and on a shelf, ransomware cannot reach it. The honest downside for a small shop is operational. You need the drive, archiving software, a verification habit, and the discipline to keep a second copy, because a single tape with a read error is a single point of failure.

Option two: cold object storage in the cloud #

Cold cloud storage flips the economics: near-zero hardware to buy, a tiny monthly storage fee, and the real bill lands when you retrieve. AWS S3 Glacier Deep Archive is the cheapest at-rest tier from a major provider at $0.00099 per GB per month in us-east-1, which is about $1 per terabyte per month (AWS S3 pricing, checked Jun 2026). The asterisks are where the money hides. Deep Archive has a 180-day minimum storage duration, charges for 40 KB of extra metadata per object, and retrieval is both slow and metered: bulk restore runs $0.0025/GB and takes up to 48 hours, while standard restore is $0.02/GB. Restoring 10 TB costs roughly $26 on bulk versus $205 on standard, and that is before egress to actually pull the bytes out of AWS (LeanOps and Usage.ai analyses of AWS rates, checked Jun 2026).

If retrieval is more than a once-in-a-blue-moon event, a flat-rate provider is often the better cold tier. Wasabi Hot Cloud Storage is $7.99/TB/month as of July 1 2026 with no egress or API charges, as long as your monthly egress stays at or under your stored volume, and it has a 90-day minimum retention (Wasabi pricing docs, checked Jun 2026). Backblaze B2 is $6/TB/month with free egress up to 3x your average monthly storage, then $0.01/GB beyond that (Backblaze, checked Jun 2026). Those flat tiers cost roughly 6-8x more per month at rest than Deep Archive, but you are never afraid to read your own data. The deeper comparison of those three as a media backend lives in B2 vs Wasabi vs plain S3, and the broader pattern of cloud retrieval fees is in what leaving actually costs.

Option three: a shelf of cheap HDDs #

The lowest-friction option for a small shop is also the oldest: write each finished project to a couple of hard drives and put them on a shelf. Even after the 2026 price spike, the sweet spot for price-per-terabyte is the 14-18 TB enterprise range, where a good deal lands under $14/TB and a great one under $11-12/TB (HDDHunt, checked Jun 2026). Retrieval is the fastest of any cold option: plug the drive in and copy. There is no drive to amortize like tape, and no egress meter like cloud.

The honest weakness is durability over time. A powered-off hard drive is not an archival medium the way tape is. Bearings, lubricant, and bit rot are real over multi-year shelf life, and a drive that has not spun in three years sometimes does not spin at all. The rule that saves you is redundancy: never one copy. Two drives per project, stored in two places, with a calendar reminder to power them up and verify a checksum once a year. Think of cheap HDDs as the affordable near-line tier for projects you might genuinely need back this year, not as the place a project goes to rest for a decade.

The retrieval tradeoff is the whole game #

Every archive decision is really a bet on how often, and how fast, you will need a project back. Get that bet wrong in the cheap direction and a routine restore becomes a budget event. The clearest example is Glacier Deep Archive: storing 100 TB there is about $100/month, which is wonderful, but if a client commissions a recut and you need all 100 TB back on a deadline, standard retrieval is roughly $2,048 in restore fees alone, plus egress, plus the wait. Bulk is far cheaper at around $256, but it can take two days. That is fine for a legal hold and painful for a Monday-morning callback.

Tape inverts the curve. The cost is mostly the up-front drive, and reads are effectively free once you own the hardware, but you wait while a cartridge loads and streams, and you maintain the gear. Cheap HDDs win on retrieval flat out, at the cost of long-term durability. The practical takeaway: match the medium to the realistic retrieval pattern of each project class. Wedding and event work you will likely never reopen can go to the cheapest cold cloud tier. A brand's master campaign assets, which get recut every quarter, do not belong behind a 48-hour wall.

A tiering plan that does not surprise you #

You do not have to choose one medium. The right answer for most small post shops is a tier ladder, and a 3-2-1 backbone underneath it: at least three copies of anything you cannot lose, on two different media, with one copy off-site. Here is a plan that scales from a solo editor to a small team.

Hot, working set: the project lives on your editing storage while it is active. Near-line, recent: on delivery, copy the project to cheap HDDs (two of them) and, if you can, to a flat-rate cloud tier like B2 or Wasabi for the off-site copy. This is where projects from the last 12-24 months sit, retrievable in minutes. Deep, cold: once a project has aged past the window where recuts are likely, push the off-site copy down to Glacier Deep Archive for about $1/TB/month, and retire it from your near-line drives. If your archive grows into the many-tens-of-terabytes range and you are reading it back often enough that cloud retrieval fees sting, that is the signal to bring LTO in-house and amortize the drive. Tag every project with its delivery date and a "review for cold" date so the migration is a scheduled chore, not a panic.

One honest note on where a mount fits, since this blog lives on a NAS-mount project. JuiceMount makes your working NAS feel like a local Finder volume for active editing; it is not an archive product, and it does not replace tape or cold cloud for projects you have retired. What it does help with is the boundary: keeping your hot and near-line tiers on storage you own, with no per-seat or per-GB tax, so the only thing you are renting is the deep-cold copy you rarely touch. For the bigger question of self-hosting versus renting the whole stack, see why on-prem is having a moment again.