What is token sustainability in AI and why it matters

As AI becomes an enterprise-wide capability, token usage is increasing rapidly. This drives both cost and energy demand, making token usage relevant to financial performance and environmental sustainability commitments.

What was once a localized R&D expense has become a distributed operational cost across the business. As token use grows, organizations need a way to ensure that consumption creates enterprise value.

This is where token sustainability in AI comes in: ensuring that token consumption consistently produces business value at scale. In simple terms, token sustainability defines what “good” looks like when AI is operating at scale: high-value outcomes, controlled costs, and measurable impact.

Token-based AI behaves like a metered utility. The more tokens are used, the more an organization pays, whether or not that usage creates value.

As token use grows, the challenge shifts from consumption to value:

• Producing meaningful outcomes
• Scaling in a controlled way
• Aligning with business and sustainability priorities

Token sustainability provides a structured way to manage this challenge.

Four capabilities that enable token sustainability in AI at scale

Token sustainability is achieved through four capabilities that enable organizations to assess, monitor and manage token usage, ensuring it consistently creates business value at scale:

• Visibility
• Efficiency
• Cost scalability
• Impact

Together, these capabilities provide the visibility, control and insight needed to assess token usage and ensure it delivers both business and sustainability value.

Visibility: understanding how tokens are used and what they cost

Visibility provides a clear view into where tokens are used, how usage is changing and what it costs.

As AI adoption expands, token usage directly influences operating costs, budget forecasting and margin visibility. Without this visibility, token consumption is difficult to track and manage effectively.

Efficiency: ensuring tokens produce value

Efficiency focuses on whether token consumption delivers meaningful results, rather than wasted or repeated work.

In unoptimized AI workflows, organizations pay for both the employee’s time and the tokens consumed to generate incorrect outputs and retry tasks. Improving efficiency reduces wasted token usage and ensures each interaction contributes to a useful outcome.

Cost optimization: managing how token costs grow

Cost optimization enables organizations to manage how token costs grow as AI adoption increases.

At small scale, token usage may go unnoticed. At enterprise scale, it becomes a significant and often unpredictable expense. This shift from fixed to variable cost makes it harder to forecast and control budgets as AI usage expands.

Impact: linking tokens to energy and emissions

Impact connects token consumption to energy demand and environmental performance.

Because tokens represent compute usage, they are directly linked to energy consumption and, in some cases, emissions. As AI scales, token usage becomes an operational factor that influences both cost and sustainability commitments.

Token sustainability enables AI to scale with control and accountability

As AI usage grows across the enterprise, token consumption becomes a critical operational factor.

Token sustainability in AI enables organizations to:

• Monitor and manage token usage
• Ensure consumption creates business value
• Control costs as usage scales
• Align AI with sustainability goals

In this way, token sustainability becomes essential for turning AI from experimentation into enterprise performance.

How EY applied token sustainability principles internally

GenAI was used to develop an iteration of this article. In accordance with EY editorial guidelines, the end product was reviewed and edited by EY professionals before publication.