Global AI eats 4.3GW, and it will triple by 2028

AI represents 4.3GW of power demand today, according to an estimate by Schneider Electric, growing at a CAGR between 26 and 36%, to reach 13.5 to 20GW by 2028.

Schneider electric global AI power report

4.3GW is about as much as Denmark consumes*.

“This growth is two to three times that of overall data center power demand CAGR of 11%,” according to the company. “One key insight is that inference loads will increase over time as more newly trained models are transitioned to production. Actual energy demand will heavily depend on technology factors including successive generations of servers, more efficient instruction sets, improved chip performance and continued AI research.”

Schneider Electric estimates	Year 2023	Year 2028
Total data center workload	54GW	90GW
AI workload	4.3GW	13.5-20GW
AI compared with total)	8%	15-20%
training vs inference workload	20% training 80% inference	15% training 85% inference
central vs edge workload	95% central 5% edge	50% central 50% edge

The figures in the table come from a report in which Schneider explains relevant attributes and trends of AI workloads, and predicts resulting data center challenges.

“Guidance to address these challenges is provided for each physical infrastructure category including power, cooling, racks, and software management,” it said.

The report, ‘AI disruption: Challenges and guidance for data center design‘ can be downloaded without payment

*un-guaranteed Electronics Weekly calculation based on 2021 figures via Wikipedia 🙂