AI’s power usage has surged, becoming a significant concern. Recent data shows AI power consumption has risen to 5.1GW globally due to widespread adoption of generative AI models. This is an increase from the previously reported 4.3GW in October 2023. As AI adoption continues, power consumption is projected to hit 13.5-20GW by 2028, marking a 26-36% compound annual growth.AI workloads will make up at least 20% of global data centre power consumption.
Check Here
“In future, we’ll see greater use of predictive analytics, #digitaltwins and new applications to fuel a new wave of business transformation. And while #AI is driving #datacentre growth, by 2028, AI workloads will make up at least 20% of global data centre power consumption.” pic.twitter.com/Cue8nQ97yk
— Schneider Electric (@SchneiderITUK) March 6, 2024
Data Centre Power Usage
AI currently accounts for around 9% of a typical data center’s 56GW power consumption. By 2028, data center usage could reach 92GW, with AI contributing 15-20%. The AI workload split, now 25:75 for training vs. inference, is expected to become more inference-heavy.
Cooling and Efficiency
Cooling remains crucial due to excess heat risks, leading to high electricity and water use. Data centers often modify water courses to manage heat. Predicting energy usage will become harder as inference workloads rise. Schneider Electric suggests transitioning to 240/415V distribution for better power density management.
Infrastructure and Efficiency
Upgrading infrastructure and improving efficiency are essential to manage power usage. This will help make cloud computing and AI workloads more sustainable. New AI hardware from NVIDIA and Intel has been designed to be more power-efficient, potentially curbing the overall power growth trend.
Table Representation:
Metric | Current Consumption | Projected Consumption by 2028 |
---|---|---|
AI Power Consumption (GW) | 4.3 GW | 13.5-20 GW |
Data Center Power Consumption (GW) | 54 GW | 90 GW |
AI as % of Data Center Power Consumption | 8% | 15-20% |
Collaboration with NVIDIA
Additionally, Schneider Electric and NVIDIA have partnered to optimize AI workloads in data centers, focusing on enhancing performance, scalability, and energy efficiency. Their collaboration aims to implement advanced AI hardware and infrastructure solutions that reduce power consumption and make data centers more sustainable. This partnership exemplifies Schneider Electric’s commitment to addressing the power consumption issues in AI-driven data centers.
Check out this article: Schneider Electric and NVIDIA collaborate to optimize AI for data center performance, scalability, and energy efficiency – https://t.co/3fQ2jHhMbc #SchneiderElectric #NVIDIA #SchneiderElectricXNVIDIA #ArtificialIntelligence #AI @melissamonico1 pic.twitter.com/0kTbhWI20o
— Maria Victoria (@vdb64) April 26, 2024