Cold Storage vs Archive Tiers: What Survives a 10-Year Gap

Storing data for the long haul looks simple on a pricing page. A tier promising to hold a terabyte for about a dollar a month is tempting, and for the right data it's a genuine bargain. The trouble starts when someone needs that data back, and the bill arrives for a number nobody modelled.

This is the gap between what cold storage and archive tiers cost to fill and what they cost to empty. Get the trade-off right and these tiers are the best value in cloud storage. Get it wrong and the surprise retrieval fees dwarf everything you saved. Here's how to tell the difference before you commit a decade of data.

Cold storage vs archiving: not the same thing

The two terms get used as synonyms, but they describe different things, and confusing them creates risk.

Cold storage is a technology tier: cheap bytes with slow retrieval and no governance layer. AWS S3 Glacier, Azure Archive and Google Coldline are cold storage. They store data inexpensively and ask nothing about what the data is or who can search it.

Archiving is a capability: governed, searchable, policy-driven, long-term data management that may use cold storage as one of its tiers. For anything with regulatory, legal or search requirements, archiving is mandatory, because cold storage alone leaves you with cheap bytes and no way to find, prove or defensibly delete them. If you only need to retain raw data you'll rarely touch, cold storage is fine. If you need to answer an auditor, you need archiving on top.

The tiers, and what they actually cost in 2026

Each major provider offers a ladder from frequent-access to deep-archive. The colder you go, the cheaper the storage and the slower and pricier the retrieval.

A few details that matter more than the per-GB number. Azure Archive is offline and needs rehydration before you can read it, capped at around 10 GiB per hour per account, which stretches large restores into days. Google Archive offers millisecond access but locks you into a 365-day minimum. AWS Glacier adds a small per-object metadata overhead that quietly bloats the bill when you archive millions of tiny files.

The trap: cheap to fill, expensive to empty

Here's the failure mode that catches teams again and again. They pick a tier by its storage price, set a lifecycle rule to dump everything old into it, and feel clever about the savings. Then the data is needed, and the retrieval bill lands.

Want a 100TB compliance archive back for an audit? On Deep Archive that's roughly $2,000 in retrieval plus a long wait, and more again in egress if it leaves the cloud. Glacier pricing is designed for data that goes in and rarely, ideally never, comes out.

The numbers around this are stark. Deep Archive at about $1 per TB per month is the cheapest place to put data, but pulling it back runs around $20 per TB plus a 12-to-48-hour wait. By contrast, providers like Wasabi and Backblaze B2 cost roughly six to seven times more per GB to store, but charge nothing for retrieval or egress. For an archive you read quarterly, those zero-retrieval providers often win on total cost, even though they look more expensive on the rate card.

The rule that decides hot vs cold

You don't have to guess. There's a clean breakeven you can apply to any dataset: estimate how much of your stored data you'll actually retrieve per month.

• Retrieve more than about 1.6 percent per month? A warmer tier usually wins on total cost, because retrieval fees overwhelm the storage savings.
• Retrieve less than that? Cold storage starts to pay off.
• Deep archive only makes sense when retrieval is genuinely rare, under roughly 1 percent of stored data per month, and you can tolerate the long thaw.

The discipline most teams skip is modelling retrieval cost alongside storage cost. Pull the access numbers for the last 12 months, not the last 12 days, and you'll usually find the right tier becomes obvious.

The costs the rate card hides

The per-GB headline is the smallest part of a real archive bill. Watch these four:

• Retrieval fees. Charged per GB to thaw data, and they scale fast on large restores.
• Egress. Moving data out of the cloud often matches or exceeds the storage cost itself, especially during audits and DR drills.
• Minimum storage durations. Delete from Deep Archive or Azure Archive before 180 days and you still pay the full 180 days. Match minimums to your retention policy or the penalties erase the savings.
• Per-object overhead and small files. Archiving millions of tiny logs or thumbnails wastes money on per-object overhead. Pack small files into larger archives first.

What about durability over ten years?

Durability is the part that's genuinely reassuring. The major providers design these tiers for very high durability, typically quoted as eleven nines, meaning the statistical chance of losing an object in a given year is vanishingly small. Data is replicated across devices and, depending on your redundancy choice, across facilities or regions.

The caveat is that durability is not the same as availability or accessibility. Your bytes surviving for a decade is one thing; being able to read them quickly, afford the retrieval, and still have the tools and formats to open them is another. A ten-year archive should be paired with a plan for how you'll actually use it later, including the occasional test restore to prove it works.

How to choose, in practice

Strip away the marketing and the decision comes down to a short rubric:

1. Map your access pattern first. How often, and how fast, will you realistically need this data? Use real historical numbers.
2. Model retrieval and egress, not just storage. Add them as separate line items and stress-test with a real transfer job.
3. Check minimum durations against your retention policy so you never pay early-deletion penalties.
4. For DR, prioritise restore speed over storage price. Match the tier to your recovery time objective, not your storage budget.
5. If the data is regulated or searchable, layer archiving on top. Cold storage alone is a compliance gap.
6. Test a restore before you depend on it. Untested cold storage is just a future disaster waiting to happen.

This kind of thinking sits in the same family as the backup discipline we cover throughout our recovery guides. If you want the ground-level version of why having a recoverable copy matters at all, our guide to recovering deleted data covers the fundamentals that any archive strategy ultimately serves.