Google shares how much energy is used for new Gemini AI prompts

Google has released a detailed methodology and initial findings on the environmental impact of its Gemini Apps AI text prompts, reporting a median energy use equivalent to less than nine seconds of TV and 0.03 grams of CO2 equivalent per prompt, significantly lower than many public estimates. The company highlights substantial efficiency gains, with median energy consumption per prompt reduced 33x and CO2 by 44x over the last year, while advocating for greater industry consistency in AI environmental measurement. However, this analysis is limited to text generation and excludes AI model training or hardware manufacturing, which contribute significantly to AI's overall environmental footprint, and Google's overall emissions increased 11% in 2024 despite data center energy reductions. This disclosure provides new data for assessing AI's operational ESG profile, though it underscores the complexity of fully quantifying the sector's environmental impact amidst rapid expansion.

Analysis

Alphabet's (GOOGL) disclosure on the environmental impact of its Gemini AI provides a granular, yet narrowly focused, view of its operational efficiency. The reported median energy use per text prompt, equivalent to under nine seconds of TV viewing, and a 33x improvement in energy consumption over the last year, are presented as evidence of significant progress in software optimization and clean energy integration. By releasing a detailed methodology that includes factors like idle chip usage and data center overhead, Google aims to position itself as a leader in ESG transparency and shape industry-wide reporting standards. However, the analysis deliberately omits the more resource-intensive processes of AI model training and non-text queries like video generation. Critically, while Google's data center energy emissions fell 12% in 2024, its overall corporate emissions rose 11%, driven by the supply chain impacts of manufacturing AI hardware and constructing new data centers. This dichotomy highlights that operational efficiency gains are currently being outpaced by the environmental costs of the AI infrastructure build-out, presenting a complex ESG profile for the company.

Key Decisions for Investors

• Investors should treat this disclosure as a positive step in ESG transparency but must recognize its limited scope, focusing on the discrepancy between improving per-query efficiency and the rising overall corporate emissions driven by the AI supply chain.
• Monitor for similar disclosures from competitors to benchmark Google's efficiency claims, as its push for a standardized methodology could confer a competitive advantage if it becomes the industry norm.
• Consider the rising overall emissions and associated capital expenditures on hardware and data centers as a material long-term risk that could attract future regulatory scrutiny and impact costs, despite the positive operational efficiency narrative.
• For ESG-mandated funds, the report provides useful data on operational carbon intensity but underscores the need to assess the full lifecycle impact of AI investments, including hardware manufacturing and construction.