Estimate the carbon emissions associated with your company's use of ChatGPT, Claude, Gemini and other AI services.
Calculate My AI EmissionsNo token data required. Get a directional estimate in just a few minutes.
Answer a few questions about the AI services your company uses. No contact information is needed until your footprint is calculated.
Both options produce an estimate. Actual usage data from your AI providers can improve accuracy over time.
Quick Estimate from your monthly AI spend, or a Detailed Estimate from plans, users and usage.
No tokens, GPUs or carbon accounting knowledge needed. Most people finish in 1–3 minutes.
We email you a link to your custom results page with your footprint, provider breakdown and confidence level.
Companies are rapidly adopting ChatGPT, Claude, Gemini and other AI services, but few understand the carbon footprint that comes with them. Exact AI emissions are hard to calculate because providers disclose different information and the energy behind a request varies by model, workload, data center and electricity source.
Training models and answering every prompt (inference) runs on GPUs and other accelerators that draw power around the clock.
Cooling, networking, storage, power conversion and backup systems add energy on top of the servers themselves.
Manufacturing GPUs and servers and constructing data centers carry lifecycle emissions before any workload runs.
The same workload can have very different emissions depending on how carbon-intensive the local electricity grid is.
Transparency matters more than false precision. When token or activity data isn't available, we start with the data companies actually have — what they spend on AI — and label the result for what it is: a directional estimate.
Calculated separately for OpenAI, Anthropic, Google and other AI services, then added together. For example, $2,000 a month is $24,000 a year × 0.134 = 3,216 kg, or 3.22 metric tons CO₂e per year.
Uses current U.S. monthly list prices for each plan. Free plans are valued at the provider's cheapest paid plan, since $0 would count real usage as zero emissions. Enterprise plans use the cost per user you enter.
It follows the Spend Tier of Watershed's open AI Emissions Framework, which maps AI service spending to U.S. Bureau of Economic Analysis sector 518200 (Data Processing, Hosting & Related Services) and applies the 2023 Open CEDA factor for that sector. Open CEDA is an environmentally extended input-output (EEIO) database, so the factor reflects the average emissions intensity of the whole sector — including upstream supply-chain impacts — rather than any single provider or model. Because price isn't the same thing as computing activity, $1 spent on AI doesn't correspond to a fixed amount of electricity, which is why the result is directional.
Spending with OpenAI, Anthropic, Google and other AI services
Upstream supply-chain emissions of the data processing and hosting sector, including hardware manufacturing and data-center construction
A separate line for each provider, so provider-specific factors can be added later
Model-, provider- or region-specific energy measurements
Actual token consumption
AI running on hardware your company owns
Water footprint
AI spending × an economic emissions factor. Low confidence, but almost any company can start today.
Provider, model and token usage × energy intensity × data-center efficiency × grid carbon intensity.
Customer-specific energy and emissions data supplied directly by AI and cloud providers.
Yes. The AI Usage Emissions Calculator is free. Answer a few questions about your company's AI providers and spending or usage, and we'll email you a link to your custom results page.
No. The Quick Estimate only needs your approximate monthly spend with each AI provider. The Detailed Estimate uses plans, number of users, usage levels and any additional API spending. Neither requires token counts.
It is a directional estimate, not an audit-grade calculation. It uses an industry-average emissions factor for data processing and hosting services, and price isn't the same thing as computing activity. The Quick Estimate is labeled low confidence; the Detailed Estimate has greater input specificity.
The calculator applies 0.134 kg CO2e per U.S. dollar, the 2023 Open CEDA factor for BEA sector 518200 (Data Processing, Hosting & Related Services) that Watershed's AI Emissions Framework recommends for its Spend Tier.
For many companies purchasing third-party AI services, the associated emissions generally fall within Scope 3 Purchased Goods and Services rather than Scope 1 or Scope 2. The right treatment depends on your organizational boundaries and accounting methodology.
Yes. Aclymate is working toward token-based estimates and, as AI and cloud providers release customer-specific energy and emissions data, provider-reported emissions. We'll update the methodology as better data becomes available.
Want to see the rest? Aclymate Explorer can estimate emissions across your company's operations, energy, travel, purchasing and supply chain.
