Why infrastructure is key to AI readiness: Insights from Think 2026

This week at Think 2026, business leaders explored a growing reality: as AI moves into the core of the enterprise, success depends on the strength and flexibility of the hybrid infrastructure behind it. The conversations made clear that AI value is built on the right architecture.

Complexity across environments, fragmented architectures, growing technical debt and rising expectations for security, compliance and resilience stand between experimentation and real, trusted outcomes at scale.

But every one of those challenges comes back to the same thing: data.

“AI is about data and that data is everywhere, so that data is essentially hybrid,” said Ric Lewis, IBM Senior Vice President of Infrastructure. “That data is either a goldmine […] or a landfill […] your architecture that you choose can influence that.”

From that starting point, Ric Lewis outlined three strategic priorities for building a strong AI foundation: putting AI at the core, enabling AI-ready data, and establishing AI-ready control. Together, these priorities help organizations create a foundation that is flexible, resilient, and built to scale with what comes next.

How enterprises manage this reality and build for it is now shaping the defining infrastructure challenge. The infrastructure leaders getting ahead are the ones treating data as the foundation their hybrid strategy depends on.

• The real bottleneck in enterprise AI isn’t models—it’s architecture. Fragmented hybrid environments quietly erode performance, trust and scalability. Aligning infrastructure, data, compute and analytics layers is now foundational to enterprise intelligence.
  
  
• Your data foundation determines your AI outcomes. AI is only as good as the data behind it and for most enterprises, that data is siloed and ungoverned. Getting data AI-ready, with the right context and governance in place, is what separates pilots from production.
  
  
• Hybrid cloud as the control plane for enterprise AI. Rather than a compromise, hybrid cloud has emerged as the operating model that allows organizations to run AI on real data, maintain sovereignty and resiliency, and scale insights reliably across environments.
  
  
• Trust is built in the stack, not added later. Trust, security and compliance are architectural outcomes, not features, embedded deep into how data and AI systems are designed.
  
  
• Delivering AI ROI means treating data, security and resilience as one system. AI value accelerates when data access, governance, cyber recovery and operational resilience are designed holistically to support production-grade intelligence.

Hybrid is the operating reality for enterprise AI. According to an IBM Institute for Business Value study, 70% of executives report that a hybrid strategy has enabled their organization to optimize costs and performance, but only 8% say their current infrastructure meets all their AI needs. That gap was at the center of Think 2026. Focusing on the following areas can help close it.

A recent Gartner report brings what’s at stake into focus: 60% of organizations are expected to drop their AI projects in 2026 over data quality concerns. CEOs are worried, too—6% say the workloads meant to generate revenue for their business are backed by data that cannot be trusted.

“AI really needs data that has referential value, quality-based information. In fact, it needs context,” said Scott Baker, VP, Marketing, IBM Storage.

These problems may sound all too familiar. Data scattered across environments, missing business context and governance treated as an afterthought. That combination is what keeps most enterprises stuck in pilots.

“We can certainly apply infrastructure that’s fit for purpose and designed for AI workloads,” said Baker. “But we still need data that’s designed for those workloads, too, with the context and quality to ensure AI performs the way it’s expected to.”

IBM takes a full stack approach to addressing the data layer, the infrastructure and the intelligence required to bring production deployments to scale. That means unifying structured, unstructured and semi-structured data across hybrid environments without requiring organizations to replatform everything they have.

“It’s this data foundation that ultimately makes the difference for an organization whether they get their AI into production or where they get stuck in this endless mode of AI pilots,” said Sam Werner, General Manager of IBM Storage.

This is where IBM Fusion and watsonx.data come in, bringing data to applications with governance enforced at the data layer. Fusion also uses intelligent data caching to bring data to where it is needed with no performance compromise.

GPU acceleration is part of the equation, too. One of the most expensive problems enterprises face today is idle GPUs, and that comes down to having storage that can push data fast enough to keep them constantly in use.
 

— Sam Werner, General Manager of IBM Storage
 

Operational resiliency planning, high availability and disaster recovery are no longer infrastructure footnotes. As AI moves deeper into core operations the stakes around availability get higher and resiliency has to be built in from the start.

“The AI needs to be up 24 by seven,” said Hillery Hunter, GM, IBM Power and CTO of IBM Infrastructure, “It’s critical to our operations.”

In a global environment, data residency is critical and sovereignty logic needs to be embedded directly into how environments are deployed, not bolted on afterward. FlashSystem  extends that to storage, providing management across an organization’s entire storage estate to keep environments resilient and compliant.
 

 — Hillery Hunter, GM, IBM Power and CTO of IBM Infrastructure
 

“Hybrid is no longer a choice,” said Lewis. As organizations navigate growing complexity, security requirements, and sovereignty concerns, hybrid is becoming a requirement not an option. Building the right hybrid architecture enables greater intentionality, consistency, and resilience, creating the conditions for AI success at the core.

At Think 2026, that message came through clearly. For enterprises running AI against real data, the architecture decisions made now will shape what is possible next. The enterprises that get there first will be the ones that build for control, flexibility and the long term.