Carbon management software provider Watershed has published a 43-page white paper proposing a token-based methodology — kilograms of carbon dioxide equivalent per million tokens, paired with electricity consumption — to give companies a practical, defensible way to estimate greenhouse gas emissions from their artificial intelligence usage. The approach leverages the token metrics engineering teams already track for cost management, creating an immediate bridge between financial operations and sustainability reporting at a moment when corporate AI adoption is far outrunning the availability of granular emissions data from model providers.
The data vacuum is real. Frontier developers such as Anthropic and OpenAI have not made emissions disclosure a priority, and even the major cloud platforms — despite their own climate commitments — offer only limited visibility. Amazon’s customer-facing dashboard does not break out AI-specific workloads, while Google and Microsoft have published technical papers on the energy, emissions, and water impact of their Gemini and Copilot systems only in recent months. Watershed’s proposal acknowledges these gaps and provides sample calculations that sustainability teams can adapt using whatever grid-intensity and vendor data they can assemble today.
For energy and sustainability professionals, the token metric is significant because it translates an abstract computational unit into a measurable environmental footprint that can be managed. Companies can now model reduction pathways — procuring matching clean energy, shifting inference to lower-carbon regions, or optimizing prompt efficiency — using the same dashboards that track API costs. This operationalizes Scope 3 accounting for AI in a way that high-level corporate pledges or voluntary vendor scorecards have not.
The white paper also outlines four concrete steps: map AI vendors and distinguish standalone models from embedded features; begin estimating with available grid and usage data; demand richer disclosures from providers; and engage in industry standard-setting. As regulatory pressure mounts — particularly in the EU and California — firms that adopt a consistent, auditable methodology now will avoid the scramble when mandatory reporting arrives. The token framework may not be perfect, but it is the first to meet sustainability teams where they already work.
Read the full report at GreenBiz.