Epoch AI Maps Rapid Growth in AI Training Compute
A New York Times visualization based on Epoch AI data highlights the rapid growth in computing power used to train advanced AI models. Epoch AI estimates that frontier language model training compute has increased about fivefold annually since 2020, doubling roughly every 5.2 months. Its analysis attributes the growth since 2018 to more chips, longer training runs and improved hardware, with these factors combining to produce about 4.27 times annual growth. Epoch AI also estimates that around 20 million AI chips were in use globally by mid-2026, measured in NVIDIA H100 equivalents. The total could double every nine months and reach 200 million by the end of 2028. The figures underline rising demand for AI computing infrastructure, but the article reports no direct cryptocurrency market development.
Neutral
The news is neutral for cryptocurrency markets because it describes AI compute growth and a data visualization, without reporting a change in crypto prices, regulation, adoption or blockchain activity. In the short term, traders may view the chip and data-centre figures as supportive of broader technology-sector sentiment, which can sometimes coincide with risk-on moves in crypto. However, that connection is indirect and is not enough to establish a clear directional signal. Longer term, continued AI infrastructure investment could influence investor demand for technology and power-related assets, but the article provides no evidence of a direct effect on crypto liquidity or fundamentals. Traders are more likely to watch broader risk appetite, rates and crypto-specific indicators than treat these estimates as a standalone catalyst.