Bitcoin miners are becoming AI power lessors, but the model still depends on execution

Bitcoin miners are not suddenly becoming AI companies, but some are being valued as power lessors for AI infrastructure. That matters because the reported opportunity is not about mining more coins. It is about monetizing power access, site infrastructure, and speed to deployment in a market where AI builders need large amounts of electricity and space.

What changed in June 2026?

Reporting on Bernstein’s June 2026 note says miners have signed at least 17 power supply agreements with major technology firms, with the deals valued at more than $110 billion. The same coverage says the agreements cover about 6 gigawatts of capacity, roughly 10% of AI data centers then under construction in the United States. Those figures come from third-party reporting, not a company announcement, but they show how quickly power-rich infrastructure has become a strategic asset.

The practical shift is that some mining sites can be repurposed into colocation or powered-shell arrangements for AI workloads. That is a different business model from pure Bitcoin mining. Instead of depending only on block rewards and mining economics, operators may earn from long-term infrastructure contracts tied to power delivery and facility use.

Who does this affect?

The immediate beneficiaries are miners with sites that can be adapted for AI or colocation use. Third-party reporting specifically highlights TeraWulf and Cipher Digital, both of which Bernstein rated Outperform. The analyst estimates cited in coverage suggest TeraWulf could reach $1.7 billion in AI revenue by 2030 and Cipher Digital $1.2 billion. Those are forecasts, not outcomes, but they explain why the market is starting to price power access as a revenue stream rather than just a sunk cost.

AI operators are affected too. Leasing capacity from miners can reduce the time and capital needed to add compute at scale. For buyers, the appeal is less about branding and more about shortening the path from contract to live infrastructure. For lenders and vendors, the question is whether the site can actually be built, financed, and commissioned on schedule.

What are the limitations and failure modes?

The biggest source-supported limitation is execution. The reporting describes long-term contracts and powered shells, but those structures still depend on financing, construction, and conversion of existing sites into usable AI capacity. The practical response is to monitor permitting, interconnection, commissioning timelines, and customer concentration, because any one of those can slow revenue recognition or reduce contracted value.

There is also demand risk. Bernstein’s thesis assumes AI infrastructure demand remains strong enough to absorb large volumes of power capacity over time. If leasing demand softens, if financing conditions tighten, or if efficiency gains reduce the need for raw power at the same pace, projected revenue can fall short. The operator owns that risk, so miners need stress tests that assume slower customer onboarding and weaker pricing.

What should operators do next?

Miners should compare the economics of self-mining against leasing power or hosting AI workloads, then test those assumptions under lower utilization and delayed buildout. AI buyers should check whether a miner-backed site can meet uptime, cooling, and delivery requirements at the needed scale. The key is to treat power as an infrastructure input with contractual and operational constraints, not as a generic commodity.

For businesses that manage recurring infrastructure payments, the lesson is familiar: new revenue lines create new billing and settlement workflows. If a company is handling cross-border or high-value payments tied to infrastructure contracts, Radom can be relevant where the fit is natural, but the decision still comes down to contract structure and cash flow discipline.

FAQ: Is this mainly a Bitcoin mining story or an AI infrastructure story?

It is both. The reporting starts with miners, but the economic driver is AI demand for power and space. The market lesson is that infrastructure owners can often monetize the same asset in more than one way.

FAQ: Should readers treat the revenue forecasts as settled facts?

No. The reported revenue figures are Bernstein estimates cited by third-party coverage. They are useful as a valuation signal, but they remain analyst projections that still depend on execution.

Historically, this was a June 2026 analyst call on a fast-moving market, not an official industry launch. What remains operationally relevant now is the underlying shift it identified: power-rich crypto infrastructure is being priced less as a mining-only asset and more as a flexible base for AI capacity.