Big Tech Claims AI Will Save the Climate. A New Report Dismantles That Narrative.
A new analysis finds 74% of AI climate benefit claims lack evidence, zero verified examples exist where ChatGPT reduced emissions, and the industry conflates traditional AI benefits with energy-intensive generative AI.
The Numbers
Climate analyst Ketan Joshi analyzed 154 AI climate benefit claims for environmental organizations. His findings revealed:
- 74% lack verifiable evidence
- Only 26% cite peer-reviewed research
- 36% cite no evidence whatsoever
- 29% reference corporate publications with no primary evidence
Most critically, the report found zero verified instances where ChatGPT, Gemini, or Copilot produced measurable emissions reductions.
The Bait-and-Switch
Tech companies conflate two distinct AI categories. Their climate benefits claims address traditional AI — predictive models, computer vision, and statistical methods — which can genuinely help with grid optimization and logistics.
However, current data center expansion focuses on generative AI: LLMs, image generators, and chatbots. Only four of 154 analyzed claims related to generative AI.
As Sasha Luccioni from Hugging Face explains: “When we talk about AI that’s relatively bad for the planet, it’s mostly generative AI and large language models.”
The Origin of Industry’s Favorite Number
The repeatedly cited claim that AI could mitigate 5–10% of global greenhouse gas emissions by 2030 originates from a Google-commissioned 2023 BCG report, which cited a 2021 BCG blog post referencing only “client experience” as its source.
The Reality of Emissions
Google:
- Emissions increased 48% since 2019
- Data center energy consumption rose 27% between 2023–2024
- Location-based Scope 2 emissions grew from 5.8 million to 11.2 million tons CO₂
Microsoft:
- Emissions up 23.4% since 2020
- Electricity consumption nearly tripled: 10.8 to 29.8 million MWh (2020–2024)
- Scope 3 emissions comprise 97% of total footprint
Industry-wide context:
- Data centers consume ~1% of global electricity currently
- Goldman Sachs projects US data center share rises from 3% (2022) to 8% (2030)
- A single ChatGPT query consumes roughly 10x more electricity than a Google search
The Playbook
Analyst Joshi identifies a “greenwashing” pattern mirroring fossil fuel industry tactics:
- Downplay negative impacts
- Promise vague future benefits
- Externalize responsibility to grid operators
- Highlight speculative avoided emissions while ignoring actual enabled emissions
Google claims its products “enabled others to reduce 26 million tons of CO₂ emissions,” yet over half derives from Google Earth Pro based on unverified corporate interviews. Meanwhile, emissions from AI applications in oil exploration and fracking logistics optimization remain unreported.
Implications for Enterprises
Scrutinize vendor claims: Demand specific, verified figures rather than projections. The probability a claim relies on solid evidence is approximately 26%.
Understand AI types: Predictive models for energy optimization differ fundamentally from energy-intensive generative AI. Ask specifically what type of AI you’re deploying and for what purpose.
Measure your footprint: Most companies lack data on their AI energy consumption before claiming sustainability benefits.
The Uncomfortable Truth
Microsoft Chief Sustainability Officer Melanie Nakagawa acknowledged in February 2025: “In 2020, Microsoft leaders referred to our sustainability goals as a ‘moonshot.’ Nearly five years later, we have had to acknowledge that the moon has gotten further away.”
The technology industry faces an irreconcilable tension: winning the AI race requires energy-intensive data centers, while demonstrating ESG leadership demands honesty about generative AI’s unsubstantiated climate claims.