- AI's growing cost problem is represented by a hidden 'EBIT-T -DA' line item for tokens, with observability proposed as the solution.
- Firms are struggling to set prices for AI services amid unpredictable token economics.
- AI costs are eroding gross margins by more than 6 percentage points, reflecting the reality many organisations are living right now.
- Observability is seen as a solution to the challenges posed by token consumption, which organisations struggle to quantify.
- Token consumption is expected to increase significantly, with predictions of 120 quadrillion tokens a month by 2030.
- Organisations are facing a gap between productivity gains and hard-dollar ROI, complicating their financial assessments.
- AI capability is now the leading criterion in platform selection, with 70% of organisations increasing observability budgets this year.
AI token costs are increasingly recognized as a critical financial metric, referred to as 'EBIT-T -DA', which includes earnings before interest, tax, tokens, depreciation, and amortization. This new line item reflects the growing impact of AI expenses on gross margins, which have reportedly eroded by more than 6 percentage points for many organizations.
The challenge lies in quantifying these token costs, as many firms struggle to establish a reliable pricing model for AI services. The phenomenon of 'token maxxing' complicates matters, leading organizations to fund behaviors without clear results. Luke Napoli
Despite the falling unit prices of tokens, overall consumption is skyrocketing, with predictions indicating a 24-fold increase in token consumption by 2030. This surge is driven by the integration of AI agents into business processes, which further complicates cost management. Will Venters from the London School of Economics noted that companies often find it challenging to manage these costs due to the unpredictable nature of AI outputs.5
To address these issues, organizations are increasingly investing in observability solutions, with 70% of firms increasing their observability budgets this year. However, only 28% currently leverage AI to connect observability data to business outcomes, highlighting a significant gap in effective cost management strategies.4
As firms navigate these complexities, the need for precise cost calculations and clear visibility into AI consumption becomes paramount.
“Enterprise spend on LLM APIs rose 36% in a single year to an average of $85,521 a month even as unit prices fell. Token consumption is forecast to increase 24 times between 2026 and 2030, to 120 quadrillion tokens a month.”
