Artificial intelligence is now embedded in nearly every part of modern business, from operations to customer experience to sustainability. But as AI adoption accelerates, so does the environmental cost of running it. Recent research shows that today’s AI systems demand significant electricity, large volumes of cooling water, and rapid cycles of energy-intensive hardware. The infrastructure powering the AI boom is not yet built for a low-carbon world.
This creates a practical challenge for sustainability leaders.
On one hand, ESG has become a data problem, and AI is increasingly essential to solving it. Regulations like California SB-253 require accurate, auditable, continuous emissions reporting across complex value chains. Manual work and scattered tools can’t keep up with the volume and granularity of data required.
On the other hand, using AI everywhere simply because it’s available, is neither efficient nor responsible – and, increasingly, nor physically feasible. Large, general-purpose models carry a heavy environmental footprint, and with AI compute becoming a scarce resource, not everyone will have access to it. What we need is a more intentional approach: AI that is precise, efficient, and designed for specific tasks rather than deployed by default.
This is the principle behind targeted AI: focused systems that deliver impact with minimal computational cost, while keeping users in control of their data. As Marie Ekeland, CEO of 2050, reminds us, AI’s resource demands mean businesses must ensure its overall effect remains positive.
The principle of targeted AI are focused systems that deliver impact with minimal computational cost.
