Why context shapes the economics of AI

As AI becomes more deeply embedded in business operations, organizations are looking beyond model capabilities and benchmark scores to new questions about economics and the value of their data. At scale, token consumption, infrastructure utilization and the cost of maintaining context increasingly determine whether AI initiatives can deliver sustainable business value.

These days, organizations creating the most value from AI may not be the ones running the largest models. They are the ones with the deepest operational knowledge and the ability to make it available when and where it’s needed. That’s because better context often means less work is required to reach a useful answer. There are fewer false starts and less redundant compute, allowing AI systems to reach useful outcomes more efficiently.

OpenAI¹ has argued that AI investments should be evaluated based on business outcomes rather than token consumption alone. That perspective suggests a broader point: context is increasingly what determines those outcomes. The more relevant context an AI system has available, the less work it must do to achieve its goal.

Whether it’s a customer support agent or a coding assistant, limited context often means more retrieval cycles, more reasoning loops and more token consumption before producing a useful result. Rich context helps AI reach the same outcome more efficiently, improving both performance and economics.

This economic lens puts a sharper focus on how data is leveraged. Many of the highest value AI applications rely on access to contextual information. While traditional analytics systems were designed to answer known questions, AI systems can reuse data, knowledge and prior work across workflows, teams and future applications. The result is that retained knowledge continues creating value long after its original purpose has been served.

Growing importance of retained knowledge

Historically, much of the data collected was tied to a specific task. Support tickets helped resolve customer problems, engineering documents supported projects already underway and reports answered immediate business questions. Once those needs were met, the long-term value of that data often received much less attention.

That logic doesn’t quite apply to AI systems. Knowledge created for one purpose can become valuable in entirely different settings. A design document written to justify a technical decision on one project may later help another team understand tradeoffs, avoid repeated mistakes or accelerate related work. What began as project documentation becomes institutional memory.

The original task may not even be the most valuable use of the data — future applications may find it has even greater purpose.

Deciding what to keep

This reality is forcing organizations to reconsider not just what data to store, but what knowledge, history and context should remain available for future employees, applications and AI systems.

Storage costs, governance and regulatory requirements still influence what should be retained, archived or discarded. AI reshapes those decisions as the value of data continues to expand. Storage infrastructure decisions that were once based primarily on cost, compliance or immediate business value now need to account for future unknown opportunities as well.

Most enterprise data already spans multicloud, on-premises and edge environments. And data gravity often makes moving that information impractical or uneconomical, which is one reason hybrid deployment models have become the norm for many organizations adopting AI.

But managing data across hybrid environments also exposes a deeper challenge: deciding what data is worth keeping and where it belongs. This is where many organizations encounter a widening gap between the value of their data and the way retention decisions are still made. Data is often treated primarily as a cost to manage, while AI turns it into an asset for future applications and opportunities.

It’s a big reason why a scalable, durable system of record matters more than ever. Enterprises need data architectures that preserve a record of institutional knowledge, operational history and AI-generated information. For many, that means embracing hybrid approaches that allow data to remain where governance, performance, sovereignty or economics require, while still making that information available to AI systems.

Economics of hard drives

As AI agents become more autonomous, they will generate and consume far more context than traditional applications. In this environment, every interaction, workflow, decision and outcome can become part of an organization’s growing body of reusable knowledge.

In fact, IDC projects annual data generation will more than triple between 2025 and 2030, growing from 218 zettabytes in 2025 to more than 718 zettabytes by 2030, with agentic AI identified as a growing source of machine-generated data.²

Even if organizations retain only a fraction of that information, the scale quickly becomes enormous. The challenge is no longer simply collecting data. Organizations must determine what knowledge should be preserved and ensure it remains accessible at the right time and in the right place. As organizations retain larger volumes of context for employees, applications and AI systems, the economics of storage are increasingly important.

Beyond lowering costs, storage economics make it practical to preserve and access knowledge at the scale AI systems require. Object storage architectures powering modern cloud platforms were built around this reality, combining software for data management with storage infrastructure capable of durably retaining data at scale. This architectural approach enables organizations to retain and reuse knowledge across employees, applications and AI systems.

The value of data no longer ends when the original task is complete. Information that remains accessible can contribute to future questions, workflows and opportunities. As AI systems increasingly rely on accumulated knowledge and context, what organizations keep may become just as important as what they create. Determining what information will matter in the future—and ensuring it remains accessible—is the only way it can continue creating value. Information that disappears never even gets a chance.

Read more: Explore how persistent context is reshaping AI systems.

Footnotes

1. https://openai.com/index/managing-ai-investments-in-agentic-era
2. IDC, Market Forecast: IDC Global DataSphere Forecast, 2026–2030, June 2026