Google’s still not giving us the full picture on AI energy use

Google's recent disclosure of 0.24 watt-hours per Gemini text query is criticized for its limited scope, omitting total query volumes and energy consumption from more intensive AI tasks like image generation, thereby obscuring the true environmental impact. This partial transparency is particularly concerning given the escalating energy demands of AI, with U.S. electricity consumption for the sector projected to reach 326 terawatt-hours annually by 2028, posing significant infrastructure and environmental challenges that warrant greater corporate disclosure.

Analysis

Google's disclosure that a typical Gemini text query consumes 0.24 watt-hours provides a specific but incomplete picture of AI's energy footprint, carrying a moderately negative sentiment. The figure's utility is limited as it represents a median for text-only queries, excluding more energy-intensive tasks like image or video generation, which the company has no immediate plans to analyze. Crucially, Google has declined to release the total number of daily queries, a key variable for assessing aggregate energy consumption, which contrasts with OpenAI's disclosure of 2.5 billion daily queries for ChatGPT. This lack of transparency obscures the full operational and environmental impact, especially when considering the broader context of AI's energy demand. The sector's projected growth is substantial, with US AI electricity needs potentially reaching 326 terawatt-hours annually by 2028. This scale is underscored by massive capital commitments, such as Google Cloud's $25 billion investment in AI on the PJM grid and the need for over two gigawatts of natural gas to power a single Meta data center. The partial disclosure by Google, while a step towards transparency, fails to address the systemic risk of escalating energy costs and carbon emissions, which are becoming material factors for the technology sector.

Key Decisions for Investors

• Investors in Alphabet (GOOG, GOOGL) should treat the company's selective AI energy disclosures as a potential ESG and operational risk, and push for more comprehensive reporting on total query volume and energy use across all modalities to accurately model future operating expenditures.
• The projected surge in AI energy demand represents a systemic, industry-wide challenge; therefore, it is prudent to evaluate technology companies' strategies for securing long-term power and managing the associated infrastructure capex, as this will likely become a key determinant of profitability.
• For those with an ESG mandate, scrutinize narrow efficiency metrics and instead focus due diligence on a company's total energy consumption, its power sourcing mix (renewables vs. fossil fuels), and its concrete plans to mitigate the significant growth in its carbon footprint.