AI Energy Consumption

At the Gartner IT Symposium/Xpo 2024, industry leaders emphasized that rising energy consumption and costs are fast becoming constraints on IT capabilities. Solutions discussed include adopting acceleration technologies, exploring microgrids, and keeping an eye on emerging energy-efficient technologies.

Thank you for reading this post, don't forget to subscribe!

With enterprise AI applications expanding, computing demands – and the energy needed to support them – are rapidly increasing. Nvidia’s CEO, Jensen Huang, highlighted this challenge, noting that advancements in traditional computing are failing to keep pace with data processing needs. “If compute demand grows exponentially while general-purpose performance stagnates, you’ll face not just cost inflation but significant energy inflation,” he said. Huang suggested that leveraging accelerated computing can mitigate some of these impacts, improving energy efficiency.

Another approach highlighted was the use of microgrids, with Gartner predicting that Fortune 500 companies will shift up to $500 billion toward such systems by 2027 to manage ongoing energy risks and AI demand. Gartner’s Daryl Plummer noted that these independent energy networks could help energy-intensive enterprises avoid dependence on strained public power grids.

Hyperscalers, including major cloud providers, are already exploring alternative power sources, such as nuclear energy, to meet escalating demands. For instance, Microsoft has announced plans to source energy from the Three Mile Island nuclear plant.

While emerging technologies like quantum, neuromorphic, and photonic computing offer the promise of significant energy efficiency, they’re still years away from maturity. Gartner analyst Frank Buytendijk predicted it will take five to ten years before these options become viable solutions. “Energy-efficient computing is on the horizon, but we have a ways to go,” he said.

Until then, enterprises will need to consider proactive strategies to manage energy risks and costs as part of their AI and IT planning.