{
  "dataset": "environment-developers",
  "title": "Developers' environmental disclosures",
  "index": "Environmental Disclosure Index",
  "url": "https://aisafetytracker.org/environment/",
  "citation": "AI Safety Tracker, \"Environmental Disclosure Index\". https://aisafetytracker.org/environment/. Retrieved 23 October 2026.",
  "licences": [
    {
      "source": "environmental-disclosure-index",
      "name": "Environmental Disclosure Index",
      "publisher": "This site",
      "licence": "MIT",
      "licence_url": "https://opensource.org/licenses/MIT",
      "attribution": "AI Safety Tracker, \"Environmental Disclosure Index\". Licensed MIT."
    }
  ],
  "generated": "2026-10-23T13:22:31+00:00",
  "retrieved": "2026-10-23T00:00:00+00:00",
  "unit": null,
  "notes": "Our own reading of primary publications, reviewed 2026-10-23. Each cell is marked disclosed, partly disclosed, reported by others, or not disclosed.",
  "count": 7,
  "records": [
    {
      "developer": "OpenAI",
      "epoch_orgs": [
        "OpenAI"
      ],
      "training": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No energy or emissions figure for any model.",
        "links": []
      },
      "sites": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "Stargate sites built with Oracle and SoftBank. The first, at Abilene, Texas, runs on Oracle Cloud Infrastructure.",
        "links": [
          {
            "label": "OpenAI",
            "url": "https://openai.com/index/five-new-stargate-sites/"
          }
        ]
      },
      "power": {
        "status": "reported",
        "status_meaning": "Reported by others, not the developer",
        "text": "Permits filed for on-site natural-gas turbines at the Abilene campus. OpenAI publishes no energy figure for any site.",
        "links": [
          {
            "label": "DCD",
            "url": "https://www.datacenterdynamics.com/en/news/natural-gas-plant-planned-for-stargate-ai-data-center-campus-report/"
          }
        ]
      },
      "inference": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "0.34 Wh for an average ChatGPT query, June 2025, stated without a method.",
        "links": [
          {
            "label": "Sam Altman",
            "url": "https://blog.samaltman.com/the-gentle-singularity"
          }
        ]
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "Google DeepMind",
      "epoch_orgs": [
        "Google DeepMind",
        "Google",
        "DeepMind"
      ],
      "training": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No energy or emissions figure for any Gemini model. The Gemma 3 model card no longer carries the pre-training figure it once gave.",
        "links": []
      },
      "sites": {
        "status": "partial",
        "status_meaning": "Partly published by the developer",
        "text": "Google's own data centres. Which sites train or serve a given model is not stated.",
        "links": [
          {
            "label": "Google",
            "url": "https://storage.googleapis.com/gweb-mobius-cdn/sustainability/uploads/7f477eb723fe0c23d03f94b90a08882b9f28187d.pdf"
          }
        ]
      },
      "power": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "65% carbon-free energy hour by hour across Google's data centres in 2025, and 100% matched with renewable purchases over the year.",
        "links": [
          {
            "label": "Google",
            "url": "https://storage.googleapis.com/gweb-mobius-cdn/sustainability/uploads/7f477eb723fe0c23d03f94b90a08882b9f28187d.pdf"
          }
        ]
      },
      "inference": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "0.24 Wh, 0.03 gCO2e and 0.26 mL of water for the median Gemini Apps text prompt, May 2025, with a published method.",
        "links": [
          {
            "label": "Google",
            "url": "https://cloud.google.com/blog/products/infrastructure/measuring-the-environmental-impact-of-ai-inference"
          }
        ]
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "Anthropic",
      "epoch_orgs": [
        "Anthropic"
      ],
      "training": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No energy or emissions figure for any model.",
        "links": []
      },
      "sites": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "Amazon's Trainium clusters including Project Rainier, Google TPUs, and from May 2026 all of SpaceX's Colossus 1 in Memphis, over 300 MW.",
        "links": [
          {
            "label": "Amazon",
            "url": "https://www.anthropic.com/news/anthropic-amazon-compute"
          },
          {
            "label": "Google",
            "url": "https://www.anthropic.com/news/google-broadcom-partnership-compute"
          },
          {
            "label": "SpaceX",
            "url": "https://www.anthropic.com/news/higher-limits-spacex"
          }
        ]
      },
      "power": {
        "status": "reported",
        "status_meaning": "Reported by others, not the developer",
        "text": "Colossus 1 holds a permit for 15 on-site gas turbines. Amazon and Google state 100% annual renewable matching for their own consumption. Anthropic publishes no energy figure.",
        "links": [
          {
            "label": "DCD",
            "url": "https://www.datacenterdynamics.com/en/news/elon-musk-xai-gas-turbines-memphis/"
          },
          {
            "label": "Amazon",
            "url": "https://www.aboutamazon.com/news/sustainability/amazon-sustainability-report-2025"
          },
          {
            "label": "Google",
            "url": "https://storage.googleapis.com/gweb-mobius-cdn/sustainability/uploads/7f477eb723fe0c23d03f94b90a08882b9f28187d.pdf"
          }
        ]
      },
      "inference": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No per-prompt figure.",
        "links": []
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "Meta",
      "epoch_orgs": [
        "Meta AI"
      ],
      "training": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "GPU hours and emissions in every Llama model card since Llama 2. Llama 3.1 405B: 8,930 tCO2e location-based.",
        "links": [
          {
            "label": "Llama 3.1 model card",
            "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md"
          }
        ]
      },
      "sites": {
        "status": "partial",
        "status_meaning": "Partly published by the developer",
        "text": "Meta's own GPU clusters. Which site trained a given model is not stated.",
        "links": [
          {
            "label": "Llama 3.1 model card",
            "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md"
          }
        ]
      },
      "power": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "100% of electricity matched with renewable energy since 2020, so market-based training emissions are reported as zero.",
        "links": [
          {
            "label": "Llama 3.1 model card",
            "url": "https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md"
          }
        ]
      },
      "inference": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No per-prompt figure.",
        "links": []
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "xAI",
      "epoch_orgs": [
        "xAI"
      ],
      "training": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No energy or emissions figure for any model.",
        "links": []
      },
      "sites": {
        "status": "reported",
        "status_meaning": "Reported by others, not the developer",
        "text": "Grok 3 was trained on Colossus, a converted factory in Memphis, Tennessee.",
        "links": [
          {
            "label": "R&D World",
            "url": "https://www.rdworldonline.com/how-xai-turned-a-factory-shell-into-an-ai-colossus-to-power-grok-3-and-beyond/"
          }
        ]
      },
      "power": {
        "status": "reported",
        "status_meaning": "Reported by others, not the developer",
        "text": "Colossus 1: 15 on-site gas turbines, 247 MW, permitted in July 2025. Around 35 ran before the permit was granted. Colossus 2: dozens of 35 MW gas turbines across the state line in Southaven, Mississippi, permitted in March 2026.",
        "links": [
          {
            "label": "DCD",
            "url": "https://www.datacenterdynamics.com/en/news/elon-musk-xai-gas-turbines-memphis/"
          },
          {
            "label": "WDAM",
            "url": "https://www.wdam.com/2026/03/11/state-approves-xais-use-gas-turbines-southaven/"
          }
        ]
      },
      "inference": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No per-prompt figure.",
        "links": []
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "DeepSeek",
      "epoch_orgs": [
        "DeepSeek"
      ],
      "training": {
        "status": "partial",
        "status_meaning": "Partly published by the developer",
        "text": "GPU hours only: 2.788 million H800 GPU hours for DeepSeek-V3. No energy or emissions figure.",
        "links": [
          {
            "label": "DeepSeek-V3 report",
            "url": "https://arxiv.org/abs/2412.19437"
          }
        ]
      },
      "sites": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "Not stated.",
        "links": []
      },
      "power": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "Not stated.",
        "links": []
      },
      "inference": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "No per-prompt figure.",
        "links": []
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    },
    {
      "developer": "Mistral AI",
      "epoch_orgs": [
        "Mistral AI"
      ],
      "training": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "Lifecycle analysis of Mistral Large 2: 20.4 ktCO2e and 281,000 m3 of water for training and 18 months of use, to January 2025.",
        "links": [
          {
            "label": "Mistral AI",
            "url": "https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai"
          }
        ]
      },
      "sites": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "Not stated in the lifecycle analysis.",
        "links": []
      },
      "power": {
        "status": "none",
        "status_meaning": "Not disclosed",
        "text": "Not stated in the lifecycle analysis.",
        "links": []
      },
      "inference": {
        "status": "disclosed",
        "status_meaning": "Published by the developer",
        "text": "1.14 gCO2e and 45 mL of water for a 400-token Le Chat response.",
        "links": [
          {
            "label": "Mistral AI",
            "url": "https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai"
          }
        ]
      },
      "reviewed": "2026-10-23",
      "source_id": "environmental-disclosure-index"
    }
  ]
}