Ethereum co-founder Vitalik Buterin shared a cost table on X on Oct. 17, 2026, comparing Bitcoin (BTC) mining with AI inference. The table estimates mining costs at $93,700 per BTC mined and AI inference costs at $26,260 per GPU-year. Buterin described mining costs as split roughly between equipment investment and electricity, while the AI estimate was dominated by equipment costs.
For Bitcoin mining, the table lists $37,600 in compute capital expenditure (capex), $40,600 for electricity used by machines and another $2,400 in electricity costs. Together, the electricity entries total $43,000, or 45.9% of the estimate. The other listed costs are $3,200 for labor, $2,500 for facility rent and $7,400 for other expenses.
The AI inference estimate assigns $19,900 to compute capex, about 76% of its total. Electricity costs are $1,050 for computers and $210 for other electricity, adding up to $1,260, or 4.8%. The remaining entries are $1,400 for labor, $900 for facility rent and $2,800 for other expenses.
One estimate is per bitcoin mined and the other per GPU-year, so their percentages do not establish which sector consumes more electricity in total. They show how the listed costs are divided within each estimate, rather than comparing equivalent computing output.
CoinGape argued that miners with flexible power contracts may reduce or redirect mining when hashprice falls, while AI operators continue to carry hardware costs after deployment. That is CoinGape’s analysis; Buterin’s post presented the cost comparison without that operational claim.
