The Quiet Repricing of Cloud Storage
Object storage was the one cloud bill nobody argued about. Falling hardware costs and new archival tiers are forcing a second look at prices that have barely moved in a decade.
For most of the cloud era, object storage occupied a strange position in the enterprise budget: large enough to matter, boring enough to ignore. A line item that grew every year was treated as a law of nature rather than a purchasing decision. That complacency is now being tested, because the cost of storing a byte has fallen faster than the price of storing a byte, and the gap between the two has become hard to defend in a procurement review.
The mechanics are not mysterious. Hard drive density continues to climb, and the drives that hyperscalers deploy today hold several times what the fleet average held a few years ago. The operational cost per terabyte of a well-run storage system — power, cooling, floor space, replication overhead — has drifted down alongside. Yet headline prices for standard object storage tiers have moved little. The majors have competed on egress fees, feature tiers and regional availability instead, each of which matters, but none of which answers the question a finance team eventually asks: why does the commodity itself cost the same as it did in 2016?
The first answer is that storage is sticky in a way compute never was. Moving a petabyte is a project, not an afternoon. Egress charges, API differences, consistency models and the sheer risk of a migration all conspire to keep data where it landed. Providers know this, and price accordingly. The second answer is that customers rarely model their storage growth. A dataset that doubles every eighteen months quietly doubles its bill, and by the time anyone notices, the archive has become load-bearing.
What is changing is the arrival of credible alternatives that attack the problem from the archive side. Deep-cold tiers from the majors themselves, priced at a fraction of standard storage with retrieval measured in hours, have matured from an awkward special case into a genuine option. Independent storage clouds with flat pricing and no egress fees have found their audience among media companies and research groups with predictable, retrieval-light workloads. The effect is not a mass migration but a negotiating position: for the first time, an enterprise can walk into a renewal conversation with a costed, credible alternative in hand.
The practical advice for a team reviewing its storage bill is unglamorous. Tier ruthlessly — most object data is read never or once, and paying hot prices for cold data is the most common waste in the category. Model egress before any migration, because moving out is where the surprise costs live. And treat storage as a negotiated contract rather than a metered utility; at scale, it always was.
The limitations of this analysis should be stated plainly. Public price lists are a poor guide to actual enterprise bills, which are shaped by discounts, commitments and private terms. Hardware costs vary by region and procurement cycle, and durability and availability guarantees are not interchangeable between providers at the same price. None of that changes the direction of travel: the byte is getting cheaper to keep, and the price of keeping it will eventually follow.
Every claim above is sourced to a document, a named person, or a record we hold. Where we could not verify a claim, we say so. Read our standards and corrections policy →