Enterprise AI spending is becoming harder to control across Asia Pacific, with about 82% of AI token spending currently not clearly linked to quantifiable business outcomes, according to new research from Accenture.
The report, The CIO’s Guide to AI Tokenomics, surveyed 750 senior executives across 17 countries, including 225 respondents from Singapore, China, Australia, Japan and India. It examines how organizations can improve visibility, governance and accountability as AI usage expands.
Accenture found that 32% of APAC companies have already exceeded their annual token budgets, while respondents expect annual token consumption to increase by 79% over the next two years.
Token spending is also becoming a significant component of enterprise AI costs. APAC companies allocate an average of 24% of their AI budgets to tokens, with 52% ranking token consumption among their top three AI cost drivers.
Lower token prices may not necessarily reduce spending. According to the research, 78% of APAC respondents said they would expand existing AI workloads or introduce new use cases if token prices fell by at least 25%, while only 5% said they would save the difference.
The research also points to gaps in AI cost management. Only 31% of APAC organizations use automated model routing to match workloads with appropriate models. Accenture estimates that about 10% of AI workloads genuinely require frontier-level capabilities, yet only 45% of AI requests are routed to the lowest-cost model capable of meeting the required quality.
Just 5% of APAC organizations use some form of chargeback, while only 18% have a dedicated TokenOps or AI economics function. Nearly half, or 47%, split responsibility between IT and finance.
Accenture identifies five practices for managing AI token economics: establishing visibility into AI usage, assigning accountability for spending, routing workloads to appropriate models, defining business value before deployment, and training employees to use AI tools effectively.
The company estimates that every $10 million in annual token spending could grow to $14.4 million within two years. Organizations adopting these cost-management disciplines could potentially avoid up to $4.4 million in additional spending while maintaining their AI initiatives.
Accenture has also launched Accenture Tokenomics, a service designed to help enterprises connect AI token consumption with measurable business outcomes while providing cost governance and optimization across AI workflows, models and agents.
“AI token costs are climbing and becoming more unpredictable as usage rates increase and more agentic workflows take flight,” said Ajoy Menon, Digital Core lead at Accenture.
The research is based on a July 2026 survey of 750 senior executives at enterprises with annual revenues exceeding $1 billion, along with 15 in-depth interviews with technology and finance leaders. Accenture also analyzed 9,368 occupational tasks from the O*NET database to estimate workloads requiring frontier-model capabilities.
