AI Tokenomics Drives Unpredictable Costs for Enterprise Buyers

Enterprises using AI services are encountering new challenges in managing costs, as token-based pricing and usage variability reshape spending patterns across the sector.

Many organizations report that usage-based pricing models and token consumption have led to unpredictable and rising expenses for AI services. These shifts have made it more difficult for buyers to anticipate and control their AI-related budgets, especially as consumption patterns fluctuate.

More than half of AI service providers now use hybrid pricing structures, which combine a fixed base fee with additional charges for usage that exceeds a set cap. This approach has become common as both buyers and sellers seek ways to balance predictability with flexibility in AI billing.

Token consumption has grown rapidly, with some enterprises observing an increase of approximately 18.6 times more tokens used per developer over a nine-month period. This trend has contributed to higher overall AI bills, even as the cost per token declines.

What the numbers show

• AI accounts for about 23% of the average enterprise cloud bill
• Roughly 26% of AI spending is estimated to be wasted
• Agentic AI tasks can use up to 1,000 times more tokens than basic tasks
• Token usage per developer rose by 18.6 times over nine months in some organizations

Agentic AI tasks, which involve more complex operations, can require up to 1,000 times more tokens than simpler inference or chat-based activities. This variation in token usage further complicates efforts to forecast and manage AI costs within large organizations.

To address these challenges, enterprises are adopting practices similar to FinOps, focusing on token-level observability, model routing, and cost governance. These measures aim to provide greater transparency and control over AI-related expenditures.

Tokenomics in enterprise AI refers to the process of measuring, attributing, and managing token consumption with detailed links to business outcomes. However, organizations continue to face difficulties due to unpredictable consumption, challenges in attributing costs, and a lack of standardized metrics across different AI vendors.

Even as the price per token drops, overall AI spending continues to climb because usage is increasing at a faster rate than price reductions. This dynamic has made AI costs a substantial and volatile component of enterprise cloud budgets.

Industry reaction

The Linux Foundation’s Tokenomics Foundation is developing open standards, specifications, and metrics for AI token usage and billing. This initiative is intended to help organizations and vendors establish more consistent and transparent approaches to AI cost management.

* This article is based on publicly available information at the time of writing.