Artificial Intelligence is poised to generate trillions in economic value, but a dose of reality is setting in. A new McKinsey report cautions that rising deployment costs could wipe out a significant portion of those gains, urging a shift in focus from
adoption to value.
The Trillion-Dollar Promise
The buzz around Artificial Intelligence often comes with staggering financial projections. According to a recent October 2026 report from McKinsey titled 'The AI value curve: The economics of applying AI to existing work', the technology has the potential to generate approximately $2.52 trillion in annual economic value globally. This value is expected to come from automating tasks and supporting existing work, fundamentally changing productivity. The report identifies sectors like financial services, education, and healthcare as having the largest potential for economic gains from AI, with financial services alone accounting for an estimated $530 billion annually. This enormous figure has fuelled a global race to invest, with companies pouring capital into AI infrastructure and software. McKinsey estimates that by 2030, around $7 trillion could be invested in AI infrastructure, a sum that requires generating about $760 billion in annual profits to justify.
A Reality Check on Costs
However, the report sounds a crucial warning: the promised returns are not guaranteed. The core issue is that many organizations are focused on 'AI adoption instead of value', a metric that overlooks the true cost of implementation. McKinsey presents a sobering scenario: a 30% increase in the effective cost of deploying AI could slash the potential global economic value by $850 billion, bringing the total down from $2.52 trillion to $1.67 trillion. That represents a potential loss of about 34%. This isn't just a global issue; the impact is localized. In the United States, for example, the same cost increase could shrink the annual value pool from $960 billion to $640 billion. It is important to note that McKinsey clarifies this is a modelled scenario, not a forecast, but it powerfully illustrates the sensitivity of AI's ROI to its underlying costs.
The Hidden Cost Iceberg
So, where are these costs coming from? They extend far beyond software licenses and initial setup fees. Many of the most significant expenses are hidden and ongoing. One of the biggest drivers is the shift to consumption-based pricing for AI models, where companies pay for every request, or 'token', processed. Inefficiently designed AI agents or workflows can lead to a massive number of requests, causing costs to spiral unexpectedly. Further expenses arise from data management, as AI systems require clean, well-organized data to function effectively, a state few large enterprises find themselves in. There are also costs associated with human supervision, verification, and the rework required when AI outputs are not perfect. These operational overheads, often unaccounted for in initial ROI calculations, are what threaten to erode the impressive top-line value projections.
The Talent and Integration Challenge
Beyond technology and data, the human element represents another major cost centre. The successful deployment of AI isn't just about installing new software; it requires a fundamental redesign of workflows and significant investment in employee skills. Companies need to support workers as they learn to use new tools, shift their activities, or even change occupations entirely. The strongest economic benefits, McKinsey notes, often come from augmenting high-skilled, higher-paid professionals rather than simply replacing lower-wage work. This means the focus is on empowering experts in fields like legal services, management, and healthcare to be more effective, which requires sophisticated integration and training. This transformation demands a strategic approach to talent and change management, which carries its own substantial, though often underestimated, price tag.
