During a break in one of my sessions at Washington, DC Climate Week, I walked into a restroom and found the argument I'd been trying to make printed across every stall: dozens of flyers detailing the energy consumption of data centers across Virginia, facilities being built at breakneck speed to meet the world's appetite for AI. I’ll admit, it was a strange place to have a moment of clarity.

Outside that restroom, the Climate Week event was full of stories of companies investing in better buildings, efficiency technologies, and the latest systems. The numbers are impressive, and exactly the stories our industry wants to hear. Yes, buildings can cut energy use by about 35% at their highest levels. But here is what those stories leave out: your building earns LEED Platinum for its intelligent energy management systems but the building manager has never once asked about the energy cost of processing the data those systems generate. And then your annual ESG report documents every kilowatt saved in your facilities, but none of the AI used to compile that report appears anywhere in the report itself.

Many companies are streamlining energy use on one side of the ledger, and exploding it on the other.

The numbers are not small. Data centers already consume roughly 1.5% of global electricity. The International Energy Agency projects that figure could approach 3% by 2030 as AI scales. A single AI query consumes significantly more energy than a standard web search. Multiply that by millions of daily enterprise interactions. A company running 50,000 AI queries a day may be drawing the energy equivalent of hundreds of additional households - off the sustainability report entirely. 

Yes, AI may reduce energy demand in ways that are harder to measure - faster workflows, less physical occupancy - but most organizations are measuring only one side of it. 

The carbon cost of AI is real and measurable. It just doesn't appear in most sustainability reports because it lives upstream: in the cloud provider, the GPU cluster, the cooling systems humming beneath the raised floors of a facility in Virginia or Oregon or Dublin. 

If AI is part of your strategy, its footprint is part of your footprint. This means companies need to start asking harder questions about it. Do we need this model? Can inference happen closer to the edge? Are our cloud providers making credible renewable energy commitments? Companies that are asking these questions will be better positioned as disclosure requirements tighten and clients begin looking into the details beyond the product and into the process.

AI is here to stay and the best way to incorporate it is to start treating it like any other business process and measure its impact accordingly. Most companies would never launch a new process, build a new facility, or launch a new product without understanding energy implications and AI should be no different. So many organizations are building AI into customer service, reporting and operations and sustainability teams need visibility into the emissions associated with them.

For most organizations, the emissions associated with cloud computing, data storage, and AI services will likely sit within Scope 3. Planning ahead is essential because the day is coming where sustainability performance will be evaluated across entire systems, not just individual assets. AI is helping organizations become more efficient and productive but when sustainability is all about understanding impact, then AI's impact belongs in that conversation too.