By Michael Skurla, Digitization Consultant and co-founder of Radix IoT

While the world is wild about artificial intelligence's (AI) transformative potential, there's a less discussed but equally crucial aspect of this revolution: the electrical power that fuels and enables technology. As AI models grow in complexity and their applications become more widespread, the electricity demand is skyrocketing, pushing our existing utility/energy infrastructure to its limits while driving innovation in power generation, processing, and grid management.

AI, particularly training large language models (LLMs) is incredibly energy intensive. Data centers, the backbone of AI development and deployment, are already major electricity hogs, and with AI's proliferation, this will only surge. This accelerated demand is leading to a renewed interest and reliance on alternative energy sources, with tech giants like Amazon and Google even investing in novel solutions such as nuclear power agreements.

While electricity demand shows no sign of leveling off, the computing industry is not just tossing this over the fence to utilities. Chip manufacturers like NVIDIA are developing revolutionary processors, such as the Blackwell architecture, designed to deliver significant performance improvements while mitigating power consumption. These advancements are crucial in ensuring that the AI revolution is sustainable. This architectural shift will be most apparent in 2025, while the power sources will become more mainstream and most likely political.

The stability and quality of our power grids are also coming under scrutiny. AI workloads can create unpredictable peak demand, further stressing the strained power grid which can potentially lead to more frequent instability. With rising power demands and the U.S. power grid facing a reliability crisis, North American Electric Reliability Corporation (NERC) has raised concerns about strains to existing power grid resources. This challenge is driving innovation in grid management, with a focus on:

1. Peak Demand Management: Strategies to predict and mitigate spikes in electricity demand caused by AI workloads.
2. Renewable Energy Integration: Increased reliance on renewable sources like solar and wind power to meet the growing demand while reducing carbon footprint.
3. Energy Storage: Advanced storage solutions to store excess energy generated during off-peak hours and release it during periods of high demand.
4. Alternative Cooling: Once a novelty, liquid cooling will become a staple for processors and their data center operators.

The AI revolution is far from just algorithms and data. None of this would be possible without electricity-which is the essential foundation that enables AI. 2025 will be an inflection point of change to support a new AI computing industry that has proven not to be a passing fad but one that will be the norm, underscoring the critical need for effective monitoring and adjustment of energy usage.

##

ABOUT THE AUTHOR

Michael Skurla is a Digitization Consultant and co-founder of Radix IoT. He has 25+ years of expertise in connected product design commercialization, focused on critical infrastructure sectors' control automation and building technology product design with Fortune 500 companies. As a speaker at global industry events on leveraging outcome-based analytics driven by data, his CAPEX savings insights empower enterprises with fact-based outcomes for ESG initiatives. An active member of ASHRAE, IES Education, IoTSF, & USGBC, and an IoT thought leader, he's appeared in CBS News, Smart Buildings Magazine, Energies Magazine, IoT Evolution, Network World, RTInsights, RFID Journal, LEDs Magazine, Critical Facilities, Oilman Magazine, IoT Playbook, IoT News, Digitalisation World, LD+A, among others.