How data requirements are reshaping AI infrastructure

As AI systems become more deeply embedded in business operations, infrastructure teams are deciding where data should reside, how it should be governed, and how it can remain accessible.

Google Cloud's latest State of AI Infrastructure report¹ highlights this view, finding that 52% of organizations now use a hybrid approach to AI, while 48% prioritize infrastructure that supports data residency, compliance controls, and local security requirements.

As cloud platforms continue to accelerate AI experimentation and deployment, organizations are accumulating larger and more diverse datasets. These findings also signal how data requirements are shaping infrastructure decisions, with greater emphasis on the need to govern, retain, and access data across multiple environments.

When Data Determines Architecture

Data residency, governance, compliance, and local security are fundamentally issues of data management and control. And for many organizations, these requirements become primary decision criteria that shape how AI infrastructure is deployed.

Different types of data are subject to different legal, regulatory, security, and operational requirements. As a result, organizations often cannot keep all of their data in a single environment. Infrastructure becomes distributed because the data itself must be governed, stored, and accessed in different ways.

This dynamic is visible in the growing focus on sovereign AI initiatives, where governments and enterprises are investing in infrastructure designed to keep critical data under local control while still supporting AI workloads.

AI Changes the Value of Data

While governance, compliance, and data management are not new requirements, AI changes the amount and the value of the data organizations retain. As that retained data becomes a source of context, knowledge, and operational insight for AI systems, data retention and governance become key considerations in how information is stored and reused to create future value.

Rather than build their own foundational models, many organizations are finding that combining commercially available models with proprietary data is more effective. As a result, control of that data becomes a key factor in determining how much value AI systems can create.

AI systems depend on accumulated context, operational history, governance controls, customer interactions, and organizational knowledge that must be retained and accessed. More than simply an input into AI systems, data has become an asset that helps determine the value those systems can create, and that changes how organizations think about architecture.

Traditionally, infrastructure teams tended to focus on where applications run. Today, they're deciding where data should reside, how long it should be retained, which policies govern it, and how it remains accessible to future AI systems.

From Workloads to Data Strategy

Google's findings are not just a cloud story, they’re also a data architecture story. Hybrid infrastructure is an architectural response to the reality that valuable data resides across multiple environments.

The industry conversation is moving beyond questions of where AI workloads should run and instead evaluating whether architecture can support the way data is created, governed, retained, and reused.

Organizations that succeed with AI will build infrastructure around the realities of their data and the value it can create. As retained data becomes a more important source of context, knowledge, and competitive advantage, the ability to preserve, govern, and reuse that data becomes increasingly important. In this view, infrastructure evolves from a platform for running workloads into the durable system of record that preserves, governs, and creates value from data.

Read more: See my recent blog post to learn how object storage has become a foundational layer for modern cloud and AI infrastructure.

Footnotes:

¹State of AI infrastructure report on hybrid cloud and GDC | Google Cloud Blog