Soaring AI costs force companies to rethink adoption as subsidies end

Artificial intelligence is becoming increasingly expensive, prompting companies to rethink their widespread adoption of the technology, according to a report from May 31, 2026. The initial 'subsidized intelligence' era, where investors absorbed costs to fuel rapid adoption, is ending as major AI firms like OpenAI and Anthropic prepare for IPOs. These companies, which have relied on venture capital and below-cost pricing, are now transitioning to profit-driven models. The end of subsidies means that businesses must pay the true cost of AI services, which have risen sharply as demand scales.

This cost surge is largely driven by AI agents, which, unlike chatbots, perform complex, multi-step tasks such as managing workflows, processing data across systems, and executing actions autonomously. Each step incurs token-based charges from underlying language models, quickly multiplying expenses. For example, an agent handling customer support might call multiple models, search databases, and generate responses—all billed per token. As enterprises deploy agents at scale, their cloud and API bills skyrocket. The report suggests that companies now must carefully evaluate ROI and prioritize high-value use cases, potentially slowing adoption of experimental or low-margin applications.