U.S. data center power demand is projected to nearly double by 2027, from 31 gigawatts to 66 gigawatts, according to Goldman Sachs research. The driver is AI infrastructure buildout. That is a hard number. Behind it is a harder question: how do companies square their public sustainability pledges with the reality of what AI consumes?
Many organizations have signed those commitments. The report notes these often do not align with their AI usage. This is not a small gap. AI workloads already account for 14% of global data center demand. That figure is expected to reach 27%. The trajectory is steep. The power requirements are massive. And the data centers housing these workloads are not going anywhere.
So edge computing enters the conversation. The idea is straightforward: process data closer to where it is generated rather than shipping everything to a central, energy-hungry facility. Less transmission loss. Less reliance on giant server farms. But the report does not present edge computing as a silver bullet. It presents it as one possible piece of a larger puzzle. The core problem remains the sheer scale of AI’s energy appetite.
The Goldman Sachs projection is worth sitting with. A jump of 35 gigawatts in two years. That is the equivalent of adding the entire electricity generation capacity of a small country, just for data centers. And this is only in the United States. Globally, the numbers would be larger. The environmental impact is not theoretical. It is measured in emissions, water used for cooling, and strain on local power grids.
Organizations face a bind. They have made promises. Many of those promises are public, tied to ESG scores and investor expectations. But AI is not a discretionary tool anymore. It is embedded in products, services, and internal operations. Companies cannot simply turn it off. The question becomes whether they can make it efficient enough to keep their pledges intact.
This is where the tension sits. Sustainability commitments often assume linear growth. AI growth is exponential. The two curves do not match. Something has to give. Either organizations revise their commitments downward, or they find technical solutions that bend the energy curve. Edge computing is one such solution. But it is not the only one, and the report does not suggest it is sufficient on its own.
What the report does suggest is urgency. The power demand is rising now. The infrastructure is being built now. Every new data center locks in years of energy consumption. Decisions made today determine the environmental footprint for a decade or more. The window to act is not wide.
Look at the numbers again. 31 gigawatts in 2025. 66 gigawatts in 2027. That is a doubling in two years. AI workloads are the primary driver. The percentage of global data center demand attributed to AI is itself projected to nearly double, from 14% to 27%. These are not abstract trends. They are concrete forecasts that shape investment, regulation, and corporate strategy.
The report frames this as a sustainability problem. It is also a business problem. Energy costs money. A data center that doubles its power consumption doubles its operating expenses, all else equal. Companies that cannot manage this will face not only environmental criticism but also financial pressure. The two pressures reinforce each other.
Where this leads is not yet clear. The report points to the need for sustainable solutions. It does not prescribe which ones will win. Edge computing is a candidate. Better chip design, more efficient cooling, and renewable energy sourcing are others. What is clear is that the status quo is not sustainable. The numbers make that plain. Something will have to change.
