Vitalik compares BTC mining and AI inference costs: electricity accounts for nearly half of the former, while the latter relies more on compute capital expenditure.

Ethereum co-founder Vitalik Buterin noted in a post that Bitcoin mining's cost structure is roughly split evenly between capital expenditure on mining hardware and electricity consumption, whereas AI inference costs are almost entirely dominated by compute capital expenditure. This highlights the difference in cost structures between the two computing businesses. For Bitcoin miners, electricity bills must be paid continuously, and profitability is jointly determined by mining rig efficiency, coin price, and changes in network difficulty; AI inference, by contrast, relies more on expensive hardware investments such as GPUs, as well as equipment utilization and computational efficiency. Meanwhile, the industrial boundaries between the two are shifting. As demand for AI compute grows, some Bitcoin mining companies have begun redirecting their existing power resources and data center facilities toward AI and high-performance computing businesses, hoping to move from pure mining to compute infrastructure services. Some mining companies are leveraging