Hidden AI Costs Are Surprising Most Businesses

TL;DR: Over half of all businesses report that their AI costs are significantly higher than they budgeted. The unexpected expenses come not from model usage itself, but from hidden needs like data governance, training, and infrastructure.
Key facts
- Category
- AI
- Impact
- High
- Published
- Source
- TechRadar
Full summary
More than half of companies say AI costs are much higher than expected, driven by hidden expenses beyond just token fees.
The initial excitement around generative AI is now meeting the hard reality of the balance sheet. According to a new report from Capgemini, highlighted by TechRadar, over half of all executives admit their organization's AI costs are significantly higher than they originally anticipated. This widespread budget overrun is causing major concern in boardrooms, especially as many companies are struggling to connect their increased spending to tangible business outcomes. The core issue is a growing disconnect between the investment in AI technology and the ability to demonstrate a clear return on that investment. While the promise of AI-driven transformation remains strong, the financial path to achieving it is proving to be much steeper and less predictable than many leaders had planned for, forcing a difficult conversation about the true price of innovation.
The primary driver of these budget overruns is not, as many might assume, the direct cost of using AI models. In fact, the price per token for many leading large language models has been steadily decreasing. Instead, the financial strain comes from a wide range of associated and often underestimated expenditures. These include the substantial costs of training and fine-tuning models on proprietary data, establishing robust data governance and security frameworks, and hiring or training specialized talent. Furthermore, scaling AI from a pilot project to a production-grade system requires significant investment in underlying infrastructure, monitoring tools, and ongoing maintenance. These ancillary costs form a large, hidden part of the AI iceberg, and organizations that fail to account for them in their initial planning are now facing the consequences of incomplete budgeting.
This trend mirrors the early days of cloud adoption, when many organizations performed a simple “lift and shift” of their on-premise infrastructure only to be shocked by unexpectedly high monthly bills. That initial cost shock eventually gave rise to the discipline of FinOps, a practice dedicated to managing and optimizing cloud spending. A similar maturation is now needed for artificial intelligence. The initial phase of rapid, widespread experimentation, driven by a fear of falling behind, is giving way to a more sober assessment of financial viability. Companies are moving from a mindset of “What can we do with AI?” to “What should we do with AI, and can we afford it?” This shift marks a critical point in the AI adoption lifecycle, where long-term success will depend as much on financial discipline as it does on technical capability.
Moving forward, business and technology leaders must adopt a more strategic and holistic approach to managing AI costs. The key is to shift focus from speculative experimentation to targeted initiatives with clearly defined, measurable business goals. This requires developing a comprehensive framework for calculating the total cost of ownership (TCO) for each AI project, encompassing everything from data preparation to long-term operational support. Organizations must get better at forecasting these hidden costs and building robust systems to track spending against specific key performance indicators. The next wave of successful AI adoption will be led not by the companies that spend the most, but by those that spend the smartest, ensuring every dollar invested in AI is directly tied to creating demonstrable value for the business.
Why it matters
For engineering and IT leaders, this highlights the need for comprehensive cost modeling beyond API fees. Unforeseen expenses in infrastructure, data pipelines, and specialized talent can derail projects and strain budgets, making robust, FinOps-style practices for AI essential for sustainable implementation.
Business impact
The inability to connect AI spending to clear business outcomes puts projects at risk of being cut. Companies are struggling to justify rising costs, which can slow innovation and cede competitive advantage to rivals who master AI cost management and prove a clear return on investment.
Related on Notifire
Related stories
Primary source: TechRadar