Sustainability must be a key focus when it comes to powering AI
By Michael Bennett, Chief Transformation Officer at Powin. (Originally published on DatacenterDynamics)
AI spend is set to double over the next few years, reaching $632 billion by 2028, and 2025 will lay the groundwork for increased investments and innovation.
While enterprise leaders find value in AI tools to drive productivity, leveraging them as assistants, search engines, and more all-in-one, the mass adoption of the technology is also driving rapid increases in energy usage in the data centers that fuel them. Take ChatGPT as an example; the tool, on average, requires 10x as much electricity to process a query compared to a Google search, straining an already overwhelmed power grid.
According to IDC, electricity costs account for 46 percent of total spending for enterprise data centers and 60 percent for service provider data centers, and this spend will only become costlier with the Department of Energy estimating data centers will consume 6.7 to 12 percent of the total US electricity by 2028 compared to 4.4 percent in 2023.
In turn, utilities, power producers, and grid operators must prepare for the subsequent power demand and costs. In 2025, mass adoption of AI technologies will continue to drive data center energy demand, spurring investments into sustainable solutions and carving out new markets for powering data centers.
The state of AI and data centers in 2025
AI remains a top priority for CEOs, with leaders citing it as the most disruptive technology over the next three years. Enterprise leaders are planning to ramp up AI deployment in 2025, but must also think about the associated impacts on the energy transition, sustainability goals, and infrastructure costs. Businesses are rightfully investing in AI to boost productivity and revenue and to remain agile and competitive. However, they may be missing the larger picture of the resources required to deploy AI tools successfully, specifically regarding energy consumption.
Given the diverse mix of generation resources on the US power grid, data center growth will also increase demand for fossil fuels. Research from Goldman Sachs predicts that incremental data center power consumption in the US alone could drive around 3.3 billion cubic feet per day of new natural gas demand by 2030, a roughly 3.5 percent daily increase from current demand. This incremental demand will require additional infrastructure investments, such as new pipeline capacity, new natural gas-fired power plants, and transmission and distribution assets.
As leaders plan for 2025, utility costs and sustainable energy consumption must be considered. An environmentally friendly approach to AI is required for organizations to balance innovation with sustainability goals.
Sustainable solutions
AI’s impact on the environment and power grid won’t be solved overnight, but market solutions exist to help address these issues and transition data centers toward a more sustainable future. By shifting to more environmentally friendly strategies and investing in sustainable solutions, data centers can reduce their impact on the environment and US power grid and optimize their operations and costs.
Battery energy storage systems (BESS) store energy from the power grid and renewable sources for later use, optimizing costs while promoting clean energy usage. Energy procurement teams supporting data center developments can leverage these solutions, as part of a larger portfolio strategy, to tap into renewables, like solar and wind, to store energy and release it during peak demand periods, helping maintain costs during peak shaving periods. BESS solutions also help balance supply and demand from the grid, and by sharing stored energy back into the grid, BESS promotes a greener grid beyond just data centers.
BESS can also be physically located on-site at data centers to provide backup power in the event of a grid outage. Traditionally, diesel generators have played a key role in backup power, but several forward-thinking developers and offtakes are moving to BESS, which have a number of desirable attributes related to sustainability.
Additional solutions include nuclear power, such as leveraging small modular reactors or re-opening closed plants to provide energy for data centers. Nuclear power is a clean energy source as well, and while developing and deploying nuclear energy is currently costly and timely, the lack of emissions from this resource has made it appealing for companies like Microsoft and Amazon.
Planning for 2025 and beyond
Leaders from all industries are prioritizing AI adoption over the next year and with increased IT spend will come increased energy usage. Efficiency is the name of the game for enterprises, and leaders should also apply this principle to powering AI use. As organizations look to partner or build data centers, prioritizing sustainable solutions to power these centers efficiently must be of equal importance.
In 2025, the energy market can anticipate increased interest and spend on innovative solutions, such as BESS, as well as additional partnerships between enterprises and alternate energy sources, like nuclear power plants. C-suite leaders can get buy-in from other stakeholders by emphasizing the cost savings and improved operations that come from investing in innovative energy solutions. Generation assets like BESS that shift reliance from the traditional grid provide opportunities for organizations to cut costs and drive reliable energy independence.
Looking at the year ahead, enterprise leaders tasked with driving operational efficiency in data centers must consider alternative energy solutions that can power AI use while also promoting cost optimization and enhanced operating performance.
