AI infrastructure

Getty Images - Jason marz

For years, “cloud-first” has been the default blueprint for enterprise modernization, promising lower costs, greater flexibility, and virtually unlimited scale. Yet research from EY found that while 65% of organizations have made strategic investments in cloud, only 32% say they are achieving the business outcomes they expected. For many organizations, that gap reflects a growing mismatch between traditional cloud assumptions and the infrastructure, governance and data requirements AI introduces. As AI places new demands on infrastructure, business leaders are confronting a more urgent question than simply whether to move to the cloud: Is a traditional cloud-first strategy still fit for purpose in an AI-driven world?

Many organizations initially migrated to the cloud for cost and flexibility. But this alone is insufficient in an AI-driven world defined by speed, data gravity, and regulatory scrutiny. At the same time, shortages of GPUs and the concentration of hyperscaler infrastructure in the US mean enterprises can no longer assume a single cloud provider will meet every operational or strategic need. These dynamics are forcing organizations to become far more intentional about workload placement across clouds, on-premises, and hybrid environments. Increasingly, the issue is not whether organizations have migrated to the cloud, but whether their infrastructure strategy is aligned with how AI actually operates.

AI changes what “good” cloud architecture looks like

Traditional cloud environments were designed for predictable enterprise demand. AI workloads are harder to predict, more compute-intensive and more sensitive to latency than many traditional enterprise applications. As a result, cloud architectures once considered modern are now the operational bottleneck, constraining performance and slowing model training.

Pressure to modernize has always reshaped enterprise technology strategy. With AI, this comes down to rethinking how infrastructure supports adoption at scale. A staggering 85% of senior business leaders fear their existing technology infrastructure cannot support AI, pushing organizations away from one-size-fits-all cloud models and toward more environments tailored to specific workloads and business needs.

Further, long-standing public cloud approaches are also challenged by this shift. According to 2026 research from Cloudian, 79% of enterprises report having already moved AI workloads from the public cloud, with 73% planning to shift additional workloads to on-premises or hybrid infrastructure within two years. As AI adoption grows, infrastructure decisions are being tightly coupled with where data lives and how quickly it can be accessed and processed. However, the risk for leaders is replacing one stringent approach with another. Repatriation is not a universal fix. While it can help organizations gain greater control over performance and flexibility, it can also introduce additional complexity around cost, talent, integration, and compliance.

Ultimately, this is not a signal that leaders should retreat from cloud strategies. The aim is to develop intentional workload strategies that meet business and governance priorities. But the larger issue is that many organizations still approach AI as a scaling challenge, not an infrastructure and governance challenge. In practice, AI success increasingly depends on whether enterprises can align their compute, data access, governance, and operational control, rather than just expand cloud capacity.

Why cloud, data and governance can no longer be separate strategies

As organizations rethink infrastructure for the AI era, leaders quickly find that scaling without strong data and governance policies introduces a cascade of challenges, such as fragmented execution and greater risk exposure.

Mary Elizabeth Porray, EY Global Vice Chair – Client Technology, believes organizations are underestimating the extent to which AI fundamentally changes enterprise operating requirements. She explains, “AI is not just another application. It’s forcing organizations to rethink their cloud strategy as a business control issue, not just a technology decision. The companies that succeed will be the ones that can scale AI effectively, without losing sight of visibility, governance, or trust.”

That need for coordinated oversight is becoming increasingly urgent as AI adoption accelerates faster than many organizations can manage the associated risks. EY research found that 78% of leaders say their AI adoption is outpacing their organization’s ability to effectively manage business risks associated with AI. To manage AI risk effectively, organizations must be able to track what AI systems exist, what data they rely on and how they interact with broader business systems and workflows – and executives agree. According to a recent survey from my company, Prosper Insights & Analytics, 32% of executives and business owners are concerned about the data visibility issues associated with AI and believe organizations need more transparency into the data AI uses.

At the same time, as AI becomes democratized across business units and teams, governance fragmentation is creating additional operational risk. Without stronger coordination, AI initiatives can quickly devolve into siloed experimentation, inconsistent controls and disconnected tooling that not only fail to deliver measurable ROI, but also make it difficult for organizations to monitor, audit and scale AI responsibly. Organizations that generate long-term value from AI will be those that effectively align infrastructure decisions with governance and business priorities, rather than treating them as separate initiatives.

Engineering AI cloud environments for trust, compliance, and control

Enterprise leaders are beginning to recognize that trust, security, and compliance must be engineered into AI cloud infrastructure from the outset, not layered on afterward.

The need for embedded governance becomes particularly acute as organizations move beyond AI assistants and toward AI agents capable of taking action across systems and workflows. A recent Prosper Insights & Analytics survey also found that 34% of business leaders and executives believe agentic AI is a good idea, underscoring how quickly organizations are moving toward more autonomous AI agents despite the governance challenges it may present.

Porray adds, “We are rapidly moving toward a future where the volume of AI agents operating within enterprises will scale significantly faster than traditional models of human oversight can effectively manage on their own. That presents a new set of demands around identity management, monitoring, auditability, and policy enforcement that fundamentally changes how we look at infrastructure. Embedding governance from the outset is essential to maintaining credibility and shareholder trust.”

These governance and compliance requirements are influencing infrastructure decisions more and more. For example, organizations may place AI workloads differently depending on the sensitivity of the data involved. A content generation tool may run in a public cloud environment, while an AI system analyzing financial records, customer data or regulated information may be deployed in a private cloud or hybrid environment to satisfy data residency, auditability, or regulatory requirements. Data residency, sovereignty, and regulatory requirements often determine where AI systems can operate, while hybrid and multi-cloud environments create new challenges around visibility and governance consistency. As a result, trust, compliance, and control are emerging as core design principles for modern AI cloud environments rather than secondary IT considerations.

The next 12 months: cloud maturity will be measured by speed and resilience

As AI reshapes how organizations design and operate cloud environments, cloud maturity will increasingly center on elasticity, availability, and speed rather than scale alone. Workload placement, data governance, and infrastructure strategy are now fundamentally interconnected decisions rather than separate technology initiatives. Organizations that can coordinate these areas effectively will be better positioned to scale AI responsibly, accelerate deployment and generate measurable ROI. Without it, organizations may find themselves unable to explain, audit or confidently govern AI-driven decisions.

In an AI-driven economy, cloud strategy is no longer simply about modernization. It is increasingly becoming a determinant of which organizations can operationalize AI at scale, and which fall behind.

Disclosure: The consumer sentiment study referenced above was conducted by my company, Prosper Insights & Analytics. This is the same dataset used by the National Retail Federation, and available from Amazon Web Services, Bloomberg, and the London Stock Exchange Group for economic benchmarking.