McKinsey flags looming AI spending surge for US agencies as subsidised deals give way to market pricing

US government agencies could face a sharp rise in artificial intelligence (AI) spending as subsidised access to AI tools expires and usage expands, with McKinsey warning that falling technology costs are not translating into lower overall bills, as per its latest report.

The report said agencies must start treating AI spending as a separate budget category and link costs to measurable outcomes to avoid unexpected bills and disruption to ongoing programmes.

"The price to run a fixed level of AI capability has fallen roughly tenfold per year and about 1,000-fold over three years. It's one of the steepest cost declines of any technology in modern history. Yet enterprise AI bills are climbing because usage is exploding at a faster rate.," the report said.


The report said introductory federal agreements in 2025 made AI tools available to government agencies at nominal prices, including some for as little as USD 1. As these arrangements expire through 2026 and 2027, agencies are expected to move towards commercial pricing, exposing them to higher costs as adoption expands.

McKinsey said the challenge is compounded by the growing use of AI agents, which can perform multiple steps to complete tasks. Such workflows routinely consume five to 30 times more tokens than a simple chatbot query.

US federal agencies reported 3,611 AI use cases in their 2025 inventory, around 69 per cent more than a year earlier. About 1,800 were already deployed or being tested, while 445 were classified as high-impact

"The potential value of federal AI contracts reached an estimated $90 billion or more. The base on which tomorrow's bills will compound is being laid right now," the report said.

The pressure is not limited to government. A McKinsey survey of 120 companies found that 93 per cent had exceeded their planned AI budgets, highlighting the difficulty organisations face in controlling costs as adoption grows.

McKinsey warned that agencies relying on pilot prices to prepare budgets could face shortfalls once commercial rates take effect. It recommended forecasting costs at post-subsidy prices, setting spending limits and alerts, and directing tasks to the least expensive AI model that meets quality requirements.

The report also called for shorter and more flexible contracts, transparent pricing and regular reviews of whether AI spending delivers measurable benefits.

"Those who wait risk surprise bills, stalled programs, and a much harder conversation with appropriators," the report said, urging government leaders to establish cost controls before subsidised arrangements expire.