Epoch AI power analysis

![](https://epoch.ai/assets/images/icons/Search-Loupe-2.svg)
Search epoch.ai


[Search](https://epoch.ai/search)

Enter a query to search for results


Previous

1

...

2

...

3

Next

[Article](https://epoch.ai/blog) [How much power will frontier AI training demand in 2030?](https://epoch.ai/blog/power-demands-of-frontier-ai-training)

paper

# How much power will frontier AI training demand in 2030?

The power required to train the largest frontier models is growing by more than 2x per year, and is on trend to reaching multiple gigawatts by 2030.

![](https://epoch.ai/assets/images/posts/2025/power-demands-of-frontier-ai-training/projected-power-growth.png)

[![](https://epoch.ai/assets/images/icons/twitter-big.svg)](https://twitter.com/intent/tweet?url=https://epoch.ai/blog/power-demands-of-frontier-ai-training&via=EpochAIResearch)[![](https://epoch.ai/assets/images/icons/linkedin.svg)](https://www.linkedin.com/sharing/share-offsite?url=https://epoch.ai/blog/power-demands-of-frontier-ai-training)[![](https://epoch.ai/assets/images/icons/speech-balloons.svg)\\
Read X thread](https://x.com/EpochAIResearch/status/1954987311365575108)

![](https://epoch.ai/assets/images/icons/citation.svg)
Cite


### Published

Aug 11, 2025

### Authors

Josh You,
David Owen

### Resources

[![](https://epoch.ai/assets/images/icons/colab.svg)\\
Source Code](https://colab.research.google.com/drive/1uF9jAWP_lnNIe7nd5_0hjgpFzE2_jKDw?usp=sharing) [![](https://epoch.ai/assets/images/icons/document-paper-angle.svg)\\
Paper](https://www.epri.com/research/products/000000003002033669)

The electrical power required to train individual frontier AI models has been growing rapidly over time, driven by the growth in total training compute and the size of training clusters. Previously, we found that the power required to train a frontier model has been [more than doubling](https://epoch.ai/data-insights/power-usage-trend) every year. If trends continue, how high could these power demands become?

In a new [white paper](https://www.epri.com/research/products/000000003002033669), “Scaling Intelligence: The Exponential Growth of AI’s Power Needs”, written in collaboration with [EPRI](https://www.epri.com/), we analyze the factors driving power growth for frontier training, and forecast this growth out to 2030. We conclude that the largest individual frontier training runs in 2030 **will likely draw 4-16 gigawatts (GW) of power, or enough to power millions of US homes**.

## Forecasting power demands using model training compute

Power demands for frontier training runs have historically grown at a rate of 2.2x per year, with the largest runs now exceeding 100 MW. This has primarily been driven by frontier training compute, which has been growing at [4-5x](https://epoch.ai/blog/training-compute-of-frontier-ai-models-grows-by-4-5x-per-year) per year.

However, translating this compute growth trend into power demand requires dividing the compute growth rate by growth rates in two m

... (truncated, 14 KB total)