The following is the Executive Summary of Trimountaine's latest paper: "Sustainable Enterprise AI Deployment". Click anywhere to request a free copy of the report.
EXECUTIVE SUMMARY:
Much of the public discourse on the climate impact of AI has focused on infrastructure development and the means with which to power this new load. There has been little discussion about the role that corporations have in responsibly deploying and using AI in a sustainable way. Partially, this is due to a gap in the accounting methodology as well as a lack of transparency on the part of the AI hyperscalers. But this doesn’t invalidate the impact that exponential AI growth is having on corporate emissions. This report attempts to unmask that impact and provide context on what corporate sustainability officers should be doing to better understand and control the impact of AI on their climate activities.
The lack of focus on enterprise AI isn’t surprising. A review of 50 Fortune 500 sustainability reports — all from companies with validated Science Based Targets and active enterprise AI deployments — found that 68% contain no AI-related emissions disclosure whatsoever. A further 28% reference AI only in the context of energy demand or operational efficiency without quantifying any GHG impact. Not a single company in the dataset explicitly discloses AI-related GHG emissions from its own enterprise use of AI.
However, corporate executives do believe that AI is already having a material impact. The Capgemini Research Institute's January 2025 survey of 2,000 senior executives found that 48% believe their AI use has already driven a rise in GHG emissions — and 42% have had to reassess their climate goals as a result. Among companies that measure their AI carbon footprint, emissions from generative AI is expected to just about double every two years. And that doesn’t look at the cascading effect of agentic AI deployments, which are expected to multiply query volumes dramatically by 2027,
By 2030 AI driven GHG emissions for non-tech companies is projected to grow from about 2-3% of total emissions today to somewhere in the 7-12% range.
Reducing the usage of enterprise AI, though, negates the cost and efficiency gains inherent in the new technology. Controlling for emissions needs to adapt to other levers. Chief Sustainability Officers (CSOs), fortunately, do have a number of tools with which they can take control. Five specific levers — model selection, cloud region and provider choice, workload timing, token and query efficiency, and modality selection — are enterprise decisions that can alter the trajectory of AI related emissions without materially changing the usage of AI within the enterprise.
The scale of enterprise AI impact is a blind spot for nearly every organization. Within the next few years, though, it will be one of the largest sources of emissions for a corporation. Taking action, today, to monitor and affect this trajectory is vitally important.
