Enterprise AI investment continues to grow, but rising operating costs are already changing how organizations use the technology. Cloudera’s latest survey, The State of Enterprise AI 2026: The New Rules of Scale, found that 55% of respondents report increased AI investment.
While increased investment may not be surprising, the survey also found that 89% of respondents have limited or redesigned AI use cases because of rising token consumption, signaling a shift in how enterprises assess AI investments. The challenge now is determining which applications can deliver enough value to justify their operating costs as usage grows.
Growing AI Investment Puts Returns Under Greater Scrutiny
Of the surveyed respondents, 43% report productivity gains that are greater than they expected. Interestingly, almost the same number (41%) report costs that are higher than they expected. There is clearly a benefit in terms of productivity from the investment, but that doesn't necessarily equate to a better financial return. In fact, more than half (58%) say their AI initiatives are generating about the same level of value as they are costing.
Consider a company using an AI assistant to take routine tasks off employees’ lists. At first, the benefits are easy to see. Employees get time back, and work gets done more efficiently. But as more people start using the assistant, token consumption rises, and so does the bill. Eventually, the company has to ask how much that saved time is actually costing them.
For enterprises expanding AI across business functions, understanding the cost of individual applications is essential to evaluating whether productivity improvements translate into financial value.
That calculation becomes harder without reliable cost data.Nearly three-quarters (73%) are operating without full confidence in their AI cost visibility. Without that visibility, enterprises may struggle to assess the financial returns of individual initiatives and decide where further investment makes sense.
Token Consumption Is Guiding How Enterprises Use AI
While overall cost visibility is important for evaluating returns, monitoring token consumption can help enterprises understand how individual applications are using resources. Yet only 19% of respondents continuously monitor token and inference costs. Without consistent tracking, organizations may have difficulty identifying changes in consumption before they significantly affect spending.
Rising token consumption is already changing how enterprises operate AI. Forty percent have reduced processing frequency, model size, context size or output volume, while 24% have paused or canceled an AI use case altogether.
Rising Costs Are Reshaping Deployment Decisions
Nearly half (45%) of respondents say their organizations have moved an AI workload to another cloud, private environment or on-premises platform in response to higher AI-related costs.
But cost is only one factor shaping workload placement decisions. Resilience is the most frequently cited reason for operating AI across multiple environments at 42%, followed by cost reduction and data sovereignty or residency requirements at 38% each. Needless to say, enterprises are juggling multiple factors when deciding where workloads make the most sense.
For enterprises looking to manage costs by changing deployment environments, the need for redesign can complicate the financial decision. Moving a workload may offer lower operating costs, but organizations must also consider the resources to adapt it to a new environment. Without accounting for that additional work, enterprises may overestimate the potential savings of relocating their AI applications.
However, the decision-making process is only half the battle. More than half (58%) of respondents say relocating an AI workload requires moderate or major redesign of the workload itself.
Building portability into AI architectures can help reduce these challenges. By designing workloads to move between environments without extensive redesign, enterprises have more flexibility to respond to changing costs and business requirements.
Considering AI Costs for the Long Term
An AI application's financial value can change as usage grows, so its initial performance is only one part of the investment decision. Enterprises must also consider what it will cost to operate an application over time and whether those expenses remain justified as adoption expands.
For enterprises planning to scale AI, the priority should be to understand which applications justify continued investment and to maintain how to adjust as costs change. That means tracking expenses as applications operate and considering how easily workloads can shift if another environment proves more cost-effective.
Download The State of Enterprise AI 2026: The New Rules of Scale to explore the full findings and learn how organizations are responding to the economic realities of AI at scale.
