Investors increasingly judge Bitcoin miners by how efficiently they produce BTC and what else their electrical and data-center infrastructure can support. CoinShares argues that the slow process of connecting new U.S. facilities to the grid has made already-powered sites unusually valuable to artificial-intelligence operators.
The scarce asset is the grid connection
A mining campus combines land, power contracts, substations, cooling and a connection capable of handling a large continuous load. Those features overlap with the needs of high-performance computing, even though an AI facility may require substantial additional networking, cooling and building work. A miner that already controls power can therefore have a scheduling advantage over a developer starting with an empty site.
CoinShares cited a data-center vacancy rate that has fallen from 10% in 2019 to about 1%, where it has remained for three years. It also pointed to roughly 2,060 gigawatts of generation capacity waiting in U.S. interconnection queues, versus about 1,300 gigawatts of installed generation. The comparison does not mean all queued projects will be built. It shows the scale of demand pressing against studies, permits, transmission and equipment availability.
A forecast, not a completed revenue shift
The firm estimated that AI could rise from roughly 30% to about 70% of listed Bitcoin miners’ revenue by year-end, potentially higher. That is a research projection rather than an industry result already recorded in audited accounts. Conversion schedules, customer contracts and energization milestones will determine how much of the modeled shift becomes revenue.
Investors should also distinguish a signed lease from recognized income. An AI customer may commit capital years before a site reaches service. Construction delays, transformer shortages, financing terms and performance obligations can alter the economics between announcement and operation.
Hybrid operators no longer track Bitcoin alone
For a pure miner, revenue is closely tied to bitcoin production, network difficulty, transaction fees and BTC’s market price. An AI hosting agreement can add longer-duration contracted income and reduce dependence on a single commodity-like activity. It can also redirect power away from mining and introduce new counterparties, capital spending and execution risk.
The result is a more complicated valuation. Hashrate and cost per bitcoin remain important, but so do megawatts under control, time to power, tenant credit quality, contract length and the cost of converting a site. Comparisons between miners become less useful when one company mainly produces BTC and another is becoming a data-center landlord.
What to verify in future announcements
Readers should look for whether a project has secured power rather than merely requested it, whether the customer agreement is binding, who funds the conversion and when revenue can begin. They should also check whether management’s AI figures describe contracted capacity, a pipeline or a broad addressable market.
Grid scarcity can create genuine option value for miners, but it does not make every mine AI-ready. The strongest cases will connect physical power rights to funded construction and credible customers while preserving transparency about how much Bitcoin exposure remains.
Sources & further reading
- Adapted from BTC-Pulse: CoinShares: U.S. Data Center Grid Bottlenecks Intensify as Bitcoin Miners’ AI Revenue Share Could Rise to 70%
- Upstream summary: Wu Blockchain
- Primary research: CoinShares market update, August 27, 2026