ETH Co-Founder Vitalik Breaks Down Bitcoin vs AI Compute Cost Structure

Ethereum co-founder Vitalik Buterin shared a “fun fact of the day” on X, posting two detailed cost tables comparing Bitcoin mining economics to AI inference.

The numbers expose a clear gap: Bitcoin mining costs split roughly 50/50 between compute capex and electricity, while AI inference is dominated almost entirely by hardware capex.

That gap gives power-rich Bitcoin mining operations a structural flexibility advantage that pure AI operators largely cannot replicate.

Ad

Ad

Vitalik’s Numbers: Bitcoin Mining Is ~46% Electricity, AI Inference Is ~5%

Buterin’s tables break down costs per unit of work. For Bitcoin mining, the approximate total cost per BTC mined sits near $93,700.

Compute capex, hardware, cooling, and power gear, accounts for roughly $37,600, or about 40%.

Electricity into the machines adds $40,600, or 43.3%. Other electricity costs lift the total electricity share to about 45.9%.

Labor accounts for $3,200 (3.4%). Facilities rent comes to $2,500 (2.7%). Other expenses, maintenance, insurance, and bandwidth, make up about $7,400 (7.9%).

AI inference per GPU-year tells a very different story. Total costs land near $26,260. Compute capex alone takes up $19,900, or roughly 76% of the total.

Electricity, by contrast, totals just $1,260 across all categories, barely 4.8% of the bill. Labor adds $1,400. Facilities rent comes in at $900. Other overheads make up around $2,800.

Buterin did not frame the comparison as a criticism of either model. He presented it as a factual contrast.

Wu Blockchain highlighted the post as a notable data point for the crypto-AI infrastructure debate, noting that Vitalik has been increasingly vocal on AI and crypto security risks.

He has also built a privacy-focused AI agent that limits identity and data leaks, signaling his deeper engagement with AI infrastructure questions.

Fun fact of the day: Bitcoin mining cost is split about 50/50 between compute capex and electricity. Meanwhile, AI inference is almost completely dominated by compute capex. pic.twitter.com/PCFKZSuNRw

— vitalik.eth (@VitalikButerin) October 17, 2026

Ad

Ad

Why Bitcoin Miners Are Structurally Positioned to Win in AI Hosting

The cost gap carries a direct strategic implication. Because Bitcoin mining electricity costs are large and variable, miners with flexible power contracts can curtail or redirect capacity when hashprice falls.

AI inference operators have no such relief. Once a GPU cluster is financed and deployed, the capex clock ticks whether or not the hardware is fully utilized.

That explains the wave of miner-to-AI pivots already underway. Core Scientific signed an AI data center deal with AMD as it shifted capacity away from mining.

Bitcoin miner MARA is targeting AI growth through a $168M EDF deal. And Anthropic signed a 20-year data center lease with Bitcoin miner TeraWulf, locking in long-term power access through mining infrastructure.

VanEck has also made the case publicly. The asset manager argued that Bitcoin miners’ AI arbitrage could be a game-changer, pointing to the same structural cost gap Buterin’s tables now quantify precisely.

More broadly, Bitcoin miners are diversifying into AI and projecting billions in revenue from hosting and colocation contracts.

Buterin’s post adds a precise, data-backed foundation to a shift the industry has been tracking for months.

Miners that control large, flexible power contracts are not just hedging, they are sitting on a structural asset that AI-only operators cannot easily replicate.

And despite the AI-Bitcoin intersection narrative, Vitalik has rejected claims that AI could crash Bitcoin’s price by 50%, reinforcing his view that the two ecosystems are more complementary than competitive.