Microsoft’s aggressive push for AI dominance has finally shattered its image as a corporate green leader. The 2025 sustainability report reveals a harsh reality: carbon emissions have surged by 25%, hitting a massive 34 million metric tons. The reason is straightforward—building the massive data centers required for large language models consumes so many physical resources that even the most polished PR campaigns about saving the planet can no longer mask the fallout.
Microsoft has essentially admitted that past attempts to appease investors through creative accounting were futile. In February, the company officially stopped purchasing "non-additional" Renewable Energy Certificates (RECs). This move is a de facto admission that old carbon offset mechanisms are little more than bookkeeping tricks that do nothing to curb the actual energy consumption of giant infrastructure. While AI hardware is being deployed at breakneck speed, sustainable technologies simply cannot keep up.
Main Drivers of the Environmental Crisis
Rapid expansion of infrastructure for training and running LLMs. Abandonment of ineffective emission offset tools like RECs. Mismatch between clean energy development and AI industry demands.
You are either building the most energy-intensive computing architecture in history, or you are keeping your word to environmentalists. In our view, the choice in favor of profits and teraflops has already been made.
The situation reflects a systemic crisis across the industry: Google reports a similar 25% spike in supply chain emissions, while Amazon consumes billions of gallons of water to cool its servers. A clear conflict of interest has emerged. On one hand, Microsoft’s ambitious goal to become carbon negative by 2030 remains on its website; on the other, the physics of the process suggests the opposite. ESG risks for shareholders are becoming a harsh operational reality that can no longer be hidden behind a screen of green certificates.