The project wrapped eight months ago. The client paid, the invoice is closed, everyone moved on. And then the email lands: “Hey! Quick one. Can we get the raw files from the shoot? We want to recut it for a new campaign.” And somewhere in your office is a drive. You think. It might be the orange one. Or the one labeled “FINAL_v2_ARCHIVE” that actually contains someone’s wedding. This is the moment archiving exists for, and most of us handle it with vibes and a shoebox.
Let’s fix that. This is the honest guide to what happens to a project after it wraps: the three real strategies (drives on a shelf, cloud cold storage, and LTO tape), what each actually costs in 2026, and the point where each one stops making sense. Fair warning, the math has changed a LOT this year, and not in the direction anyone expected.
Last updated: July 2026. Some links below are affiliate links. Buy through them and it helps keep the site running, at no extra cost to you. I only point at gear I’d be comfortable handing to a friend.
The quick verdict
If you just want the answer: most freelancers should archive to two big external drives (one home, one elsewhere) with checksums, plus AWS Glacier Deep Archive as the cheap third copy for anything irreplaceable. Small studios sitting on 100+ terabytes of wrapped projects are the ones who should price out an LTO-9 tape deck, which suddenly looks like a bargain now that hard drive prices went to the moon. Nobody reading this should buy LTO-10. We’ll get there.
The rule that makes everything else make sense
Before the gear: 3-2-1. Three copies of anything you care about, on two different types of media, one of them offsite. It was coined by a photographer (Peter Krogh, in The DAM Book) and it’s survived because it maps to how things actually die: a drive fails (copy two saves you), ransomware eats your NAS (the offline copy saves you), your office floods (the offsite copy saves you). Every strategy below is really just a different way of buying those three copies. An archive with one copy isn’t an archive. It’s a countdown.
And the elephant: hard drives got expensive. AI datacenters bought out essentially all of 2026’s drive production, prices are up around 50% since September, and the 24TB external that was $280 in January is $680 today. I know. The reason this matters here: the “just buy more drives” strategy quietly doubled in price, while tape media ($5 per terabyte) and cloud cold storage ($1 per terabyte per month) didn’t move. The whole decision tree tilted this year.
Strategy 1: Drives on a shelf (the right answer for most freelancers)
Unfashionable, unglamorous, works. Two copies of every wrapped project on two separate drives, stored in two separate places. The gear is boring and the discipline is everything, so let’s do gear fast and discipline properly.
The workhorse is a big desktop external. Right now the best dollar-per-terabyte in that category is the Seagate Expansion 24TB at about $680 (B&H has it cheaper than Amazon at the moment, with US stock). Yes, that hurts, see above, and yes it was $280 in January. It includes Seagate’s Rescue recovery service, and one drive swallows roughly two years of a typical freelancer’s wrapped projects. Buy two. That’s the whole trick. One lives at your place, one lives at the studio, your parents’ house, a friend’s rack, anywhere that doesn’t share your plumbing and your burglar.
Once you’re archiving regularly, bare internal drives plus a dock beats buying enclosures forever. The Sabrent dual-bay USB-C dock is $47 and takes two naked 3.5-inch drives at once, so your A and B copies write simultaneously. (It’s 5Gbps, and before anyone asks: no spinning drive on earth saturates 5Gbps. A 10Gbps dock buys you nothing here.) Naked drives then need armor: ORICO’s anti-static cases are about $21 for five, they stack like tiny lunchboxes, and they replace the shoebox-of-loose-drives setup that kills archives via static and dings.
Now the discipline, because this is where shelf archives actually die. First: checksums, not drag-and-drop. A checksum is a fingerprint computed from the file’s actual bits; verify it after the copy and you have proof the archive is real, not a 97%-copied folder that LOOKS fine in Finder. From my DIT days this was the whole religion: Hedge’s OffShoot on Mac or the free TeraCopy on Windows do it automatically. Second: label drives with what’s ON them and WHEN (a spreadsheet or free app like NeoFinder for the catalog). Third: spin every archive drive up once a year and re-verify. Unpowered drives are estimated to hold data for years, but estimates are doing heavy lifting in that sentence, and the honest practice is to migrate the whole archive to fresh media every 3 to 5 years. The shelf is cheap; the calendar reminders are the actual product.
Strategy 2: Cloud cold storage (the offsite copy you don’t have to drive anywhere)
Cloud archive pricing splits into two philosophies, and picking wrong costs real money, so here’s both.
AWS Glacier Deep Archive is the cheapest storage that exists, full stop: $0.00099 per gigabyte per month, which is a dollar per terabyte. Forty terabytes of career back catalog costs $40 a month to keep. The catch is that it’s cheap in one direction. Files have a 180-day minimum, restores take 12 to 48 hours before your download even starts, and between retrieval fees and AWS’s egress pricing, pulling a 10TB project back down costs roughly a thousand dollars. Deep Archive is where projects go that you’re 95% sure you’ll never touch again… it’s insurance, priced like insurance, with a deductible like insurance. (It’s also genuinely hostile to set up if you’ve never touched AWS. Budget an afternoon and a tutorial.)
Backblaze B2 is the anti-Glacier: $6.95 per terabyte per month, files available instantly, and downloads are free up to three times your stored total per month. Seven times the storage price, but when the client email from the intro arrives, the restore costs zero dollars and starts immediately instead of costing $1,000 and starting Thursday. It also plugs into tools editors actually use (Hedge, Archiware, rclone). B2 is the right cloud when there’s a realistic chance the project comes back. (Wasabi exists too at $7.99/TB with a 90-day minimum; B2 simply beats it for this use case, so I’ll save you the comparison shopping.)
The gotcha nobody prices in: your upload pipe. Ten terabytes at a typical 100Mbps upload is nine or ten DAYS of continuous saturation. On a 500Mbps fiber upload it’s still two. Cloud archiving is a strategy you run project-by-project as they wrap, not a thing you do to five years of backlog in a weekend.
Strategy 3: LTO tape (the studio answer, and suddenly a bargain)
Tape sounds like a punchline until you see the media price. An LTO-9 cartridge holds 18 terabytes and costs $90. That’s five dollars a terabyte, offline, immune to ransomware by physics, rated for 30 years of shelf life. While drive prices went up 50% this year, that number didn’t move. Hollywood, broadcast, and every facility with an archive room runs on this stuff, and it’s not nostalgia.
The gate is the deck. The sane buy for a Mac-based shop is the mLogic mTape LTO-9, a desktop Thunderbolt 3 drive at $6,599 that includes Hedge’s Canister software (the $399 app that makes tape drag-and-drop simple instead of a command-line hazing ritual). It writes about a terabyte an hour. The math: two tape copies of everything cost about $10 per terabyte in media, versus roughly $60 per terabyte for two shelf drives at today’s prices. Run the break-even and the deck pays for itself somewhere around 130 terabytes of archived material. Before the drive shortage that number was 300-plus, which is why LTO used to be a facilities-only conversation and now every studio owner with a full NAS should at least do the napkin math. Full disclosure: my own archive has never justified a deck, but the DIT and facility side of my world runs on LTO, and the workflow is genuinely solved at this point.
“Wait, should I get LTO-10 instead? Bigger number!” No. Emphatically no. LTO-10 broke the format’s oldest promise: every previous generation could read the tapes of the two generations before it, and LTO-10 reads nothing but itself. Not one LTO-9 tape, ever. The only desktop Thunderbolt LTO-10 drive costs $19,049 (not a typo), and the new 40TB LTO-10 cartridges that started shipping this year only feed that ecosystem. LTO-9 is the generation for humans in 2026, its media just got cheaper per terabyte, and the drives will be sold and supported for years. Buy the mature generation. Let enterprises beta test the new one.
Who should NOT buy any of this: if your wrapped-project pile is under 50 terabytes, an LTO deck never pays for itself. Two drives and a Deep Archive bucket beat it on every axis at that scale. The deck is a 100TB-plus purchase, or a “client contracts require LTO deliverables” purchase (yes, that’s a thing in commercial and broadcast work, and it instantly settles the question).
The three strategies, side by side
| Strategy | Upfront cost | Cost per TB | Restore speed | Best for | Weak spot |
|---|---|---|---|---|---|
| Two drives on shelves | ~$1,360 (two 24TB) | ~$60/TB (two copies) | Minutes (plug it in) | Freelancers, <50TB total | Discipline, media refresh every 3-5 yrs |
| Glacier Deep Archive | $0 | $1/TB/month | 12-48 hrs + ~$100/TB to pull | The “never touch again” third copy | Restore cost, upload time, AWS setup |
| Backblaze B2 | $0 | $6.95/TB/month | Instant, free egress (3x) | Projects likely to come back | Rent forever, adds up past ~50TB |
| LTO-9 tape | ~$6,599 deck | ~$5/TB per copy | Hours (mount + restore) | Studios, 100TB+, contract deliverables | Upfront cost, LTO-10 migration someday |
What to archive (and what to delete, guilt-free)
Keep: camera originals, audio, the project file, exported masters, and anything licensed or unrepeatable (graphics packages, client-supplied assets). Delete without ceremony: render files, proxies, cache, optimized media, and autosaves… every one of those regenerates from the originals. On a typical project that’s 20 to 40% of the folder gone before you copy a byte. If your project folders are organized enough that you can actually tell those apart at wrap time, congratulations, and if they’re not, that’s exactly what our folder structure exists to fix. Wondering how many terabytes a project even is? The video storage calculator does the GB-per-hour math for every codec.
FAQ
How long do hard drives really last on a shelf?
Estimates for an unpowered drive in decent conditions run 9 to 20 years, but that’s an estimate, not a warranty, and the research is thinner than anyone would like. The practice that removes the question: spin up and verify yearly, migrate to fresh media every 3 to 5 years. Treat the refresh cycle as insurance, not paranoia.
Is the LTO “30-year archival life” real?
It’s real the way an EPA mileage number is real: achieved under ideal conditions, meaning stable temperature around 61 to 77°F and moderate humidity. A tape in a climate-controlled closet will outlive every hard drive you own. A tape in a garage in Phoenix will not see 30. Store them like you’d store film.
What about Blu-ray or M-DISC?
At 25 to 100GB a disc, optical is a rounding error against video project sizes. Archiving one 4TB doc project means burning and cataloging 40+ discs. Lovely for family photos and final masters; not a video archive strategy.
Can’t I just leave everything on my NAS?
The NAS is the working copy, and RAID is not a backup (deletion, ransomware, and power supplies do not respect parity). A NAS full of dead projects is also the most expensive shelf you own, especially at 2026 drive prices… wrapping projects OFF the array is what frees those bays for work that pays. Archive is where projects go so the NAS can breathe.
Where this fits in your workflow
The working-storage side of this story lives in the NAS guide and the new NAS hard drives guide, the RAID levels are decoded in the RAID explainer, and the checksum habit starts on set with a proper offload workflow. Size your next archive with the storage calculator.
And that’s it! As always, what you see here is what works for me, my clients, and my crew. Three copies, two media, one offsite, checksums on everything, and a calendar reminder to verify. No strategy on this page is exciting, which is exactly the quality you want in an archive. Questions, or an archiving horror story that’ll make the rest of us feel better? Comments are open. The rest of our guides live on the Gear Guides page.
