I’m a member of GAVINE MOONEY, an Australian energy expert
An Analysis of Outstanding AI Data Center Energy Consumption!!

Power growth around the world is mainly due to energy conversion, and power growth in the IT industry is incidental!

Data center computing is up 550% with energy consumption up just 6%. Boom! 💥

How much will data center energy consumption increase in the future? It won’t be as much as you think.

Forecasts for data center energy consumption have often been inflated because they are based on simple analyses leading to conventional wisdom that increasing demand for data center services will also increase global energy consumption.

This overlooks the strong offsetting effect of simultaneous energy efficiency improvements.

Global data center computing instances increased by 550% from 2010 to 2018, but global data center energy consumption increased by only 6% from 194 TWh/year to 205 TWh/year.

This means that energy intensity in global data centers, expressed as “energy usage per computing instance,” has decreased by 20% annually since 2010.

From 2010 to 2018, data center energy consumption was driven by:
More than 6x increase in 🔼 workloads and computing instances
🔼 IP traffic increases by more than 10x
Approximately 25x storage capacity

However, this was largely offset by improvements in other areas such as:
🔽 4x reduction in electricity usage per calculation
🔽 Storage Drive Energy Usage (kWh/TB) ~9x reduction
🔽 5x fewer servers per workload

Data center efficiency (PUE), which is the total amount of energy used by the data center divided by the energy used by the IT equipment, also increased steadily.

Data center energy consumption will definitely increase with the advent of Generative AI and LLM, but how much will increase is still unknown.