Artificial intelligence promises to generate trillions in economic value. But a closer look at recent analyses from consulting giant McKinsey reveals a harder truth: for many firms, the staggering costs of AI are spiralling, and the returns are not keeping
up.
The Multi-Trillion Dollar Promise
It’s hard to ignore the headline numbers. McKinsey's research estimates that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion in value to the global economy annually. This immense potential comes from AI's ability to boost productivity, enhance creativity, and streamline complex processes across nearly every industry, from financial services to healthcare. This projection has fuelled a global investment frenzy, with boardrooms everywhere pushing to adopt AI to avoid being left behind. The promise is a new frontier of productivity and economic growth, making AI investment seem less like a choice and more like a necessity.
A Reality Check on Runaway Costs
Beneath the optimistic forecasts, however, a more troubling picture is emerging. McKinsey's own data highlights that AI costs are expanding rapidly and unpredictably. A recent report titled 'The AI Value Curve' modelled a scenario where a 30% rise in the effective cost of AI deployment could slash the potential annual economic gains by a staggering $850 billion. The problem is that many organisations are measuring their progress by AI adoption, not by the actual value it creates. These costs are not just about software licenses. They include massive spending on computing power, the high salaries of specialised talent, the extensive work of cleaning and preparing data, and the complex challenge of integrating new AI systems with legacy technology. One survey revealed that around 20% of organisations felt their use of AI was constrained by these operating costs.
The Scaling and Strategy Trap
One of the biggest hurdles isn't the technology itself, but the strategy behind it. Many companies get stuck in 'pilot purgatory', where small-scale AI experiments succeed but never get scaled to the enterprise level where they could generate significant returns. According to McKinsey, a key reason for this failure is that the limiting factor is often the organisation's ability to absorb change. There is a tendency to simply use powerful AI to automate existing, sometimes low-value, tasks rather than fundamentally reimagining business processes. The firm stresses that true value comes from using AI to augment highly skilled professionals, not just replace manual work. Companies that do succeed are those that build a repeatable capability to transform their efforts into new ways of working, which is a significant organisational challenge.
The Unpredictable Economics of 'Tokenomics'
The rise of generative AI has introduced a new, volatile cost variable: tokens. Every interaction with a large language model consumes tokens, and as usage scales, these costs can spiral out of control. McKinsey notes that the cost to complete a single task using an AI agent can vary by as much as 30 times depending on the process, creating massive budget uncertainty. This has led to the emergence of 'AI tokenomics'—a discipline similar to financial operations (FinOps) for cloud computing, aimed at managing and forecasting AI usage to maximize ROI. Some companies have even exhausted their annual AI budgets in just a few months, forcing difficult renegotiations. This unpredictability makes it clear that simply deploying the technology without rigorous cost management is a recipe for financial strain.
Finding a Smarter Path to AI Value
The message from these findings is not that AI is a bad investment, but that a blind one is doomed to fail. To ensure gains outweigh the costs, leaders must shift their focus from adoption to value. This begins with a clearer assessment of where AI can solve specific, high value business problems, rather than a broad, tech first approach. It requires investing in a solid data foundation before scaling, as poor data quality is a primary roadblock. Most importantly, it demands a strategic plan that accounts for the true endtoend costs—from infrastructure and talent to the ongoing management of 'tokenomics'. The companies that win with AI will be those that treat it not as a magical black box, but as a powerful business tool that requires disciplined strategy and sharp economic oversight.
