Bitcoin Miners Are Moonlighting as AI Landlords

Miners are renting power to AI — here’s how

Guess what? The folks who used to just crank up ASIC rigs and mine Bitcoin are now also leasing out their power, land, and hookups to thirsty AI customers. Seven of the top ten miners by hashrate are already pulling revenue from AI or high-performance computing (HPC) projects, and the rest are lining up to do the same. Think of it as exahash by day, GPU hotel by night.

Big headline deal: a miner signed long-term hosting agreements that imply roughly $1.85 million per megawatt per year in headline revenue — a tidy benchmark others are waving around when courting AI tenants. Other miners have struck similar deals: multi-year arrangements with cloud AI providers, additional capacity deals for GPU hosting, and commercial AI clouds using H100/H200 or DGX gear. Meanwhile, some companies are buying land and power capacity specifically to build AI campuses.

The mechanics are simple. Miners’ sites already have transformers, substations, fibre, and plenty of energized pads. That infrastructure is valuable to AI operators who want high-density halls for GPUs. Instead of only running ASICs at full tilt, miners can sign long-term contracts with GPU customers and create a second, steadier revenue stream.

Why it matters — money, power, and the future of hashrate

Let’s do a little back-of-the-napkin math so the story isn’t all buzzwords. One megawatt of modern ASICs (roughly 17 J/TH) equates to about 0.059 exahash per second of network share. With recent prices and network conditions, that slice of the network has translated to about $1.0 to $1.6 million per MW per year in gross mining revenue before paying for power and operations. In plain English: running ASICs still pays — but some AI hosting deals are paying more.

That’s the reason miners are tempted. Contracted AI hosting offers predictable, long-duration cash flows, which investors love more than the volatility of mining income and fee spikes. When a miner locks in multi-year GPU contracts, it buys optionality: if Bitcoin soars, they can always redeploy capacity; if markets get grim, they still have contracted income coming in.

Macro trends give this move fuel. Analysts project U.S. data-center electricity demand to climb substantially over the next decade as AI workloads explode, and regional grid operators expect record peak demand with data centers listed among the primary drivers. Utilities are already shifting capital toward these load requests, increasing multi-year investment plans to support the pipeline of customer-backed projects.

There are, of course, speed bumps. Interconnection timelines, the availability of peaking plants and gas turbines, transformer lead times, and GPU inventory cycles all limit how fast miners can flip megawatts from ASICs to GPUs. GPU supply is a wildcard: fresh Blackwell-class chips and allocation by hyperscalers will determine how quickly AI halls can fill up.

On the mining side, a sustained jump in on-chain fees could narrow the revenue gap — a few tenths of a Bitcoin per block in extra fees can meaningfully raise per-MW mining revenue. But unless fees jump dramatically and permanently, the $1.5–$2.0 million per MW per year range for high-density AI hosting is shaping up as a practical benchmark in the U.S.

Strategically, this pivot could slow the growth rate of global hashrate into 2026 if sizable new power allocations flow to GPUs instead of ASICs. That doesn’t mean fresh capital won’t enter mining — it will — but it does make raw hashrate a weaker shorthand for a company’s value. Equity investors are now tracking contracted AI megawatts, dollars per MW per year, and utility capex plans alongside ASIC deliveries.

What to watch next: new filings that show contracted AI megawatts and dollars-per-MW disclosures, updates to utility capital spending plans and interconnection queues, regional load forecasts, and short-term averages of bitcoin fees versus subsidy. Those metrics will tell you how fast campuses energize, how much power moves from miners to AI tenants, and whether the per-MW revenue gap widens or narrows.

Bottom line: large miners are quietly evolving into mixed-use data-campus operators. Some will still be pure-play ASIC farms, but many are becoming landlords for AI — and if the checks keep clearing, that diversification might be the best survival play when the crypto weather turns sour.