When AI budgets balloon: What enterprises are learning in 2026

AI was supposed to deliver efficiency. That was the promise that fueled aggressive investments across enterprises over the past few years. But a growing body of evidence tells a different story—one that’s starting to feel familiar for IT leaders. 

A recent article in Inc. highlights a stark reality: Many companies are seeing AI costs rise much faster than expected, with budgets expanding rather than shrinking. In fact, Gartner® predicts worldwide spending will reach $2.52 trillion this year. And a KPMG survey finds that many organizations have shifted to a usage-based AI model, leaving many shocked at the cost of AI.  

 

 New research from the Flexera 2026 State of ITAM Report and the Flexera 2026 AI Pulse Report reinforces this trend, revealing that organizations are still struggling to gain visibility and control over AI-driven spending. 

 

 

Taken together, these signals point to a clear conclusion: AI isn’t inherently cost-saving. Without disciplined management, it can quickly become one of the biggest drivers of IT spend. 

The promise vs. reality of AI cost savings 

For many organizations, the business case for AI seemed straightforward. Automating tasks, improving productivity and reducing manual effort should translate into lower costs. 

The reality is more complex. 

The Inc. article describes how enterprises are encountering unexpected expenses tied to AI—compute costs, data storage, model usage fees, integration work and the ongoing operational burden of deploying and scaling AI workloads. What’s especially notable is that these costs often increase as adoption grows, rather than stabilize. 

In other words, success can actually make the problem worse. The more teams use AI, the more infrastructure, tokens and services they consume.  

This aligns closely with what Flexera’s research has been surfacing: AI isn’t a one-time investment. It introduces an ongoing consumption model that behaves much more like cloud spending—dynamic, scalable and difficult to predict. 

AI is following the cloud cost curve 

If this all sounds familiar, it should. AI is effectively retracing the same path organizations experienced with cloud adoption. Early expectations of cost savings often gave way to unexpected overruns once usage scaled. 

The same patterns are emerging 

Across both Flexera reports, several consistent themes stand out: 

1. Consumption drives cost growth: AI usage tends to expand quickly across teams. As more business units experiment with models, tools and agents, consumption increases—and so do costs. What starts as a pilot can rapidly become an enterprise-wide expense

2. Visibility is limited: Organizations still struggle to see where AI spend is happening. Shadow AI—tools and services adopted outside centralized IT—adds another layer of complexity. Without clear visibility, it becomes difficult to allocate costs or enforce accountability

 

 

3. Pricing models are complex: Layered pricing models make forecasting and optimization far more difficult than traditional software licensing. AI pricing structures can include: 

• • Token-based consumption 

• • API usage fees 

• • Infrastructure costs tied to GPUs or specialized hardware 

• • Storage and data processing fees
    

4. Optimization lags behind adoption: Teams prioritize speed and experimentation over efficiency. As a result, organizations often scale AI workloads before they fully understand how to manage them cost-effectively. This delay creates a gap between adoption and governance

Why AI costs are harder to control 

AI introduces a fundamentally different cost dynamic compared to traditional IT assets.

 

AI workloads span multiple layers, and each layer contributes to total cost, making it harder to identify where optimization efforts will have the biggest impact. These layers include:  

• Data pipelines 

• Training and inference infrastructure 

• Third-party APIs and services 

• Integration into applications and workflows 

Usage is unpredictable 

Unlike fixed licenses, AI consumption fluctuates based on: 

• User demand 

• Model complexity 

• Frequency of requests 

• Data volume 

This variability makes it difficult to set reliable budgets. 

Efficiency gains don’t always translate to savings 

AI can improve productivity, but that doesn’t automatically reduce spend. In many cases, efficiency gains lead to increased usage, which offsets or even exceeds the original savings. For example, a team that automates a process with AI may run that process more frequently or expand its scope—driving up consumption. 

The hidden drivers of AI budget overruns 

The KPMG report calls attention to several cost drivers that organizations often underestimate. Flexera’s research reinforces these insights with broader enterprise data. 

• Infrastructure costs: AI workloads often require specialized infrastructure, including GPUs and high-performance computing environments. These resources come at a premium and can scale quickly with demand 

• Data costs: AI depends on large volumes of data. Storage, processing and movement of that data can significantly increase overall spend 

• Model usage and APIs: Many organizations rely on external AI services with usage-based pricing. As adoption grows, API calls and token consumption can become a major expense category 

• Integration and operational overhead: Deploying AI into production environments requires ongoing engineering, monitoring and maintenance. These operational costs are easy to overlook during initial planning 

What Flexera’s 2026 research adds to the picture 

As AI adoption accelerates, a clearer picture is starting to emerge—one where rising costs, limited visibility and evolving governance challenges are shaping how organizations move from experimentation to long-term management. 

AI spend is becoming a governance challenge 

Organizations aren’t just dealing with higher costs; they’re struggling to manage them. AI spend often falls outside traditional IT asset management frameworks, creating gaps in tracking and accountability. 

ITAM is evolving to include AI 

The scope of IT asset management is expanding to cover, AI tools and services, consumption-based pricing models, and data and infrastructure dependencies. This evolution reflects a broader shift: AI is no longer experimental. It’s becoming a core part of the IT estate. 

FinOps principles are extending to AI 

Just as FinOps emerged to manage cloud costs, similar practices are now being applied to AI: 

• Monitoring usage and spend 

• Allocating costs to business units 

• Identifying optimization opportunities 

However, many organizations are still early in this journey. 

Moving from experimentation to discipline 

Organizations need to transition from rapid AI adoption to disciplined AI management. 

• Start with visibility: You can’t control what you can’t see. You need a clear view of where AI is being used, which teams are consuming resources and how costs are distributed across services. This requires integrating AI into existing asset and spend management frameworks 

• Establish accountability: AI costs should be tied to specific teams or use cases. When business units understand the financial impact of their usage, they’re more likely to optimize 

• Prioritize efficiency early: Optimization shouldn’t be an afterthought. Consider model selection and sizing, usage patterns and frequency, as well as opportunities to reduce unnecessary consumption 

• Treat AI as an ongoing investment: AI isn’t a one-time purchase. It’s an ongoing operational expense that requires continuous monitoring and adjustment 

The bigger lesson: Innovation without governance is expensive 

The excitement around AI is well deserved. It’s already transforming how organizations operate and compete. But the cost story is a reminder that innovation alone isn’t enough. Without governance, visibility and accountability, even the most promising technologies can become financial liabilities. 

For enterprise leaders, the takeaway is simple but important: AI doesn’t reduce costs by default. It amplifies whatever systems you have in place—good or bad. Organizations that invest in strong management practices will be better positioned to capture AI’s value without letting budgets spiral out of control. 

 

Discover the Flexera 2026 State of ITAM Report