Nearly 60% of Organizations Overshoot AI Budgets as Businesses Shift Toward Multi-Model Strategies, Capgemini Report Finds

Rising artificial intelligence costs have become a major concern for 53% of senior executives, according to a new snapshot report by the Capgemini Research Institute based on a September 2026 survey of 500 CXO-level leaders across major global enterprises.

Over the past 12 months, 59% of organizations reported that their AI-related expenditure was significantly higher than budgeted—exceeding expectations by an average of 24%, while 21% overshot their budgets by more than 30%.

This lack of financial visibility has triggered tangible consequences, with two-thirds of overspending organizations experiencing reduced free cash flow or cuts to discretionary spending.

The report highlights a critical blind spot in enterprise AI governance: while half of executives admit AI budgets are approved without full cost visibility, only a third can confidently estimate indirect expenses such as human oversight, rework, and compliance.

Consequently, just 56% of organizations are confident in forecasting future costs as AI adoption scales, prompting CFOs to take a more active role in investment approvals.

Volker Darius, CEO of Capgemini Invent, emphasized that companies must stop treating AI costs merely as a technology or procurement issue and instead manage AI as a portfolio of economic outcomes—balancing cost, quality, and business value.

To rein in escalating expenses, 53% of organizations have tightened spending controls or scaled back initiatives, while over three in five executives plan to adopt a multi-model AI strategy within the next 12 months.

This approach involves dynamically routing tasks across lower-cost proprietary, specialized, or open-weight models—which nearly 60% of firms intend to expand—based on cost, latency, and risk requirements.

Furthermore, enterprises are overhauling how they measure return on investment. Although 57% of executives cite proving business value as their biggest funding hurdle, only one in five currently tracks “AI Yield”—a metric measuring the verified business value of an AI outcome minus its full safe-delivery cost.

Over the next six to 12 months, organizations plan to shift away from simple budget tracking toward outcome-driven metrics, prioritizing first-pass success rates and cost per completed task to ensure sustainable, profitable AI scaling.