science & techOur World in Data

The compute behind leading AI systems has doubled roughly every six months since 2010

the gist

Training compute measures the total calculation used to build a notable AI system. Since 2010 this amount has doubled roughly every six months, far faster than the pace of Moore's law for computer chips. The growth has been driven by deep learning, which needs huge amounts of processing and data to improve. Bigger training runs have been a major reason recent AI systems have become so much more capable.

so what

This trend explains why only a few organisations can afford to build the largest models, since costs rise with compute. It also makes the growth in AI capability easier to track and forecast.

Our World in Data · CC BY 4.0

Image: Our World in Data