How Much Does a Petabyte Cost: On-premises vs. Cloud Storage Pricing

A petabyte of data, stored the same way, for the same five years, can cost you $490,000 or it can cost you $1.5 million. That's not a typo, and it's not two different workloads. It's the exact same 1PB, holding your typical backups, media files, or long-term archives. The only variable is where you put it and who you pay to keep it there. 

We wanted to know why the gap is that wide, and what's actually driving the sticker price on both sides of the on-premises vs. cloud debate. So we built a real 5-year Total Cost of Ownership (TCO) model, priced it out across on-premises infrastructure and two public cloud providers, and broke down every line item, from hardware and admin time to egress fees and API calls, that goes into storing 1PB of data. 

The short version: cloud storage isn't automatically the cheap option, and on-premises isn't automatically the expensive one. It depends entirely on who you're buying from and what they're charging you for after the fact. Before we get into the actual math, it helps to look at the cloud storage services market and what's driving its growth. That context is what makes the numbers below worth caring about, not as an abstract exercise, but as a decision more organizations are actively facing right now.

Market context: growth of cloud storage services

Shared, multitenant cloud storage keeps expanding, in usage and in raw capacity. IDC's most recent forecast puts the IaaS storage market at a 17.6% five-year CAGR, reaching $136 billion by 2029. That's a big number, and it points to real room for purpose-built cloud storage services to keep growing. Three trends explain why. 

1. Enterprise data growth: Enterprises typically forecast 30-40% annual growth in their storage footprint. More devices, more sources generating data, and now AI agents and applications. The result is compounding capacity requirements. 

2. Ongoing data and IT budget migrations to the cloud: Organizations keep moving applications and data off traditional on-premises environments for security, performance, scalability, or accessibility. But we're still early: Gartner estimates just 50% of legacy, customized application workloads will have moved to the cloud by 2027. There's no universal rule for which workloads belong where; every migration path looks different. What's consistent is that nearly every organization has run or is running this evaluation on some part of its infrastructure. 

Budgets are moving even faster than workloads. Gartner expects cloud-based applications to reach 87% of total application software spend by 2029. Spending is outpacing migration, which tells you where the momentum actually is. 

3. Increased cost-efficiency: Cloud lets organizations skip capital spending on hardware refresh cycles, get out of perpetual software licensing, and reduce the internal headcount needed to babysit infrastructure. The upshot: IT budgets stretch further. 

That third trend is what this article is focused on. Cost is the top consideration in most cloud infrastructure decisions, and getting an accurate handle on TCO is often the deciding factor. Most organizations building a business case for cloud adoption start here. What follows is a directionally accurate cost and TCO breakdown, based on Wasabi and WD's own analysis. 

How much is a petabyte? How much enterprise data is there?  

Before we dive into cost, some context around stored capacity is essential to help understand how much data enterprises are creating and storing. Let’s start with the baseline: 1 petabyte. That’s equal to 1,000 terabytes (using decimal), or 1 million gigabytes. In more tangible terms, 1 GB of storage is equivalent to 50-100 high resolution digital photos, approximately 10 hours of MP3 audio at 320kbps, or approximately 15-30 minutes of compressed 1080p video.  

Zooming out, IDC estimates the “Global StorageSphere” (the total installed base of stored capacity worldwide) will reach 13 zettabytes in 2026. There are 1 million petabytes in a zettabyte. How do we generate so much data? Some perspective: as of 2022, the US Library of Congress stored 21 PB of digital content, or 0.000021 ZB. Meanwhile, companies like Meta operate hundreds of exabyte-scale storage clusters that serve all of its external and internal products, including Facebook, Instagram, Reality Labs, Meta AI, Ads, Data Warehouse, and internal Databases.  

Based on current capacity and growth rates of today’s real-world use cases, organizations will be responsible for storing tens of millions of petabytes of capacity over the next few years. 

How much does a petabyte cost? 

Now that we’ve established a baseline around enterprise storage capacity, let’s discuss the cost of this storage, and how costs can differ significantly across environments (e.g., public cloud vs. on-premises) and storage providers. For this exercise, we’ll stick with a 1 PB stored capacity baseline and make a few assumptions regarding the storage scenario, including: 

1. Capacity needs to be stored for five years (to match the typical length of an enterprise storage system contract) 

2. Data is primarily unstructured and will be placed in scalable object storage 

3. Data is associated with a “secondary” storage use case (e.g., backup, media/video storage, web content, user-generated content, long-term archive) and does not require extreme levels of performance. In other words, workloads that are a good fit for HDD-based storage with average levels of usage in terms of access/performance. 

With these parameters in place, we can make a few assumptions and develop an accurate, 5-year TCO for 1 PB of stored data within today’s two leading storage deployment environments: public cloud and on-premises. 

Let’s start with on-premises: