A Seagate Technology study shows that 99% of IT managers expect AI to increase storage needs in their organisations over the next three years. 32% expect growth of more than 50% in storage requirements as a result of AI. But only 38% of organisations say they are ready to meet the increased demands.
What’s happening in practice
The study was based on a survey of 2,712 technology decision-makers in organisations in the US, China, India, the UK, Germany, France, and Japan, conducted in May–June 2026. The gap between expectation and readiness is critical: nearly every organisation understands it will need more storage, but fewer than 40% have built infrastructure that can support it.
This isn’t just a question of buying disks. AI systems generate data at several layers: the data that comes in (documents, images, calls), the data the model generates (answers, logs, embeddings), and data for audit and improvement. Each of these layers requires separate storage planning.
Why it matters beyond the headline
For an Israeli organisation planning a custom AI system, this gap means you need to plan data and storage infrastructure as part of the project from the start, not as something you deal with later. A system that processes financial documents, for example, needs to store the original documents, the extracted text, the answers the model gave, and the logs for audit purposes.
If the infrastructure isn’t suitable, the system will stall or cost far more than planned. This isn’t something you can easily fix after the system is already in production and working with clients.
What to do this week
If you’re in the planning stage of an AI system, ask whoever is building the system how much data it will generate per month, where it will be stored, and how much it will cost when the system grows by two or five times. If you’re already working with a system, check actual storage volume against the forecast you received at the start.
Properly built systems plan storage in layers: hot data for fast access, cold data for archive, and a clear deletion policy. This isn’t just a technical question, it’s also a question of ongoing operational costs. If you want to understand how to plan a system that won’t stall because of data, read about AI systems built for organisations with suitable infrastructure from the start.
Sources
Frequently asked questions
How much storage should I plan for an AI system that processes documents?
It depends on document volume and what the system stores. A system processing 1,000 documents per month can generate between 10GB and 100GB, depending on whether it stores only text or also embeddings and full logs. Plan for at least three times the expected volume in the first year.
Do cloud providers handle AI storage automatically?
No. Cloud providers supply storage, but you need to define which data is saved, for how long, and in which storage tier. Without planning, costs can grow very quickly, especially if the system stores everything in the expensive tier.