Enterprise AI is becoming deeply embedded in business workflows, increasing the need to sustain and scale it across the enterprise. The challenge is making that growth work as costs, governance requirements and workload demands evolve.
The opportunity is still significant. Enterprises want AI to drive greater productivity and business value, while maintaining the cost, control and flexibility to support it at scale. AI’s growing role in the business puts greater pressure on enterprises to maintain that balance.
To understand how enterprises are navigating these challenges, Cloudera examined these dynamics in its latest report, State of Enterprise AI 2026: The New Rules of Scale, surveying 1,500 IT leaders and decision-makers worldwide. The findings reveal what happens as AI ambitions meet the realities of operating at scale, and how enterprises are adapting as a result.
The Economics of AI Are Changing the Equation
During the initial wave of AI adoption, enterprises increased their investments with the expectation that those investments would deliver meaningful returns. That investment is continuing, with 55% of respondents reporting increased AI spending. But the economics become more complicated as usage grows. While 58% say the value generated by their AI initiatives is approximately equal to their cost, less than one-quarter (23%) say they generate more value than they cost.
One factor changing that equation is the rising cost of tokens and inference. Nearly nine in 10 enterprises (89%) have limited or redesigned AI use cases as a result of rising costs, yet only 19% continuously monitor token and inference costs. As usage increases, those variable costs can quickly change the economics of a use case and force enterprises to reconsider how, and even whether, it should continue to operate.
That makes AI economics more than a budgeting issue. Cost pressures are already influencing application design and workload placement. Forty-five percent of respondents have moved a workload to another cloud, private environment or on-premises platform, while 40% have reduced processing frequency, model size, context size, or output volume. Another 24% have paused or canceled an AI use case altogether.
The economics of AI can change quickly, and enterprises need enough visibility and flexibility to respond when they do. With AI investment continuing to grow, enterprises need greater visibility into where costs originate and the flexibility to adjust workloads to keep returns sustainable.
Production Is Where AI Gets Put to the Test
The true measure of success is when AI is put to work within an enterprise's operations. Production can expose challenges that are harder to see during development. Nearly every enterprise surveyed (98%) has pulled back an AI deployment at least once because of failure, hallucinations or unexpected behavior. For 68%, these pullbacks happen more frequently, showing how production can expose issues that are harder to see during development.
Those productivity gains put the high rate of deployment pullbacks in a different context. Forty-three percent report greater-than-expected productivity, nearly twice the 23% reporting lower-than-expected gains. Enterprises are seeing meaningful upside even as they encounter setbacks and adjust along the way.
AI’s production track record is proving more mixed than a binary measure of success suggests. Pulling back a deployment, identifying a problem and making changes can all be part of operating AI at scale. As enterprises gain more experience with AI in production, reassessing and adjusting deployments will become an important part of sustaining their value.
Governance Becomes Part of the Cost of Scale
Wider AI adoption is raising the bar as to how enterprises govern and control it. Three-quarters (75%) of respondents cite governance and compliance as a constraint on scaling AI, while 55% say governance and compliance costs have grown faster than expected–the highest of any AI-related cost measured.
Enterprises appear confident in the controls they have today, with 76% saying they can easily apply consistent data governance, security and access policies across their AI environments. The harder test comes with scale, when those policies must hold across a growing range of AI deployments.
Finding the right balance is still a work in progress. Just over half (54%) believe their organization has struck an effective balance between the speed of AI adoption and appropriate governance. The remaining respondents are split evenly: 23% believe they are moving too cautiously, while 23% say adoption is outpacing their governance and controls. The focus is on an approach that provides the right level of oversight as AI continues to scale.
Where AI Runs Can’t Be a Fixed Decision
Enterprise AI workloads have to meet a range of requirements–from cost and performance to governance and regulatory obligations. Those requirements rarely stay the same. Economics can shift, production needs can evolve, and regulations can change, making where a workload runs an increasingly important business decision.
For many enterprises, that already means operating across multiple environments. Eighty-four percent say residency, sovereignty or jurisdiction-specific requirements influence workload placement at least occasionally. Resilience is the most frequently cited reason for operating across environments at 42%, followed by sovereignty and residency requirements and cost reduction at 38% each, and proximity to data at 37%.
Having multiple environments does not necessarily mean enterprises can move freely between them. Fifty-eight percent say relocating an AI workload still requires moderate or major redesign. As requirements change, enterprises need the flexibility to reconsider where workloads run without repeatedly rebuilding the applications and controls that support them.
Building for What Changes Next
If enterprises could restart their AI journeys, nearly half (49%) would strengthen governance from the beginning; 46% would invest in data quality earlier; and 45% would modernize infrastructure sooner. Another 45% would invest more in employee training. The foundations that enterprises are prioritizing today are often the ones they wish they had put in place earlier.
What works for an AI workload today may need to change as its requirements and economics evolve. Enterprises need the flexibility to respond without repeatedly rebuilding the data, governance and infrastructure foundations that support their AI initiatives.
Sustainable AI at scale depends on a flexible approach that adapts as the technology and business needs change. With consistent governance and control across environments, organizations can give themselves more room to adapt to what comes next while continuing to pursue value from AI.
Explore the State of Enterprise AI 2026: The New Rules of Scale report to learn more about how enterprises are navigating the realities of AI at scale.
