The Use Case Series, Issue 6 of 8: Bitcoin Mining Is Now AI Infrastructure With a Side of Hashrate

Bitcoin mining has been criticized for years on energy grounds. The criticism misses what mining actually is. Mining is an interruptible electricity buyer. Rigs can ramp consumption up or down in seconds, which makes them structurally useful to grid operators dealing with renewable intermittency, stranded generation, and peak-demand management.

Paraguay routes excess hydroelectric capacity into mining rather than exporting it at low rates. Texas grid operators use mining as a demand-response asset under ERCOT programs. Russia is relocating mining operations to northern regions to monetize otherwise idle energy infrastructure. The pattern is the same in each case. Mining converts surplus or stranded electricity that would otherwise be wasted into a globally fungible digital asset.

Mining sites turned out to be the most valuable real estate of the AI build-out.

The 2025 and 2026 evolution of this story is the convergence with artificial intelligence. The same sites, power contracts, cooling infrastructure, and balance sheets that support Bitcoin mining now host AI training and inference workloads. The hyperscale AI build-out requires massive amounts of power, dense cooling capacity, and grid relationships that take years to assemble. Bitcoin miners had assembled exactly that infrastructure, for a different purpose, by the time AI compute demand surged.

More than $70 billion in AI and high-performance computing contracts have been signed by publicly listed Bitcoin miners. Some operators may derive up to 70 percent of revenue from AI by the end of 2026. The transition is being financed through senior secured notes backed by Big Tech tenant credit, which has reshaped the cost of capital for the entire mining sector.

MARA, Core Scientific, IREN, and Cipher are not pure-play miners anymore. They are hybrid infrastructure businesses.

The public mining companies that survived the 2024 halving have completed an identity transition that few observers anticipated. They are no longer pure-play hashrate producers. They are hybrid infrastructure businesses sitting on land, power, cooling, and connectivity that AI hyperscalers urgently need.

MARA partnered with Starwood in early 2026 to deliver approximately one gigawatt of near-term IT capacity, with a roadmap to scale beyond 2.5 gigawatts. Core Scientific and IREN have moved aggressively into hosting agreements with AI tenants. Cipher Mining has issued senior secured notes at 7.125 percent, a financing cost that would have been unavailable to a pure-play miner two years ago. The credit market has effectively converted miner lease risk into Big Tech credit risk, which is a structural shift in how the sector is valued.

The mining sector now has three revenue streams, and only one of them is Bitcoin.

The first revenue stream is block rewards plus transaction fees, the original miner economic model. The second is AI compute leasing, which carries longer-duration contracts, higher gross margins per megawatt, and Big-Tech-grade counterparty credit. The third is grid services revenue, including demand response, capacity payments, and curtailment credits, which monetizes the interruptibility property that makes mining unique among large industrial loads.

The combination is more resilient than any of the three streams individually. A miner with a balanced revenue mix is less exposed to Bitcoin price drawdowns, less exposed to AI capex cycles, and less exposed to grid policy changes than a single-stream operator would be. That diversification is one of the reasons the credit market has been willing to fund the transition at attractive rates.

Bitcoin's security is now subsidized by the AI economy.

The third-order effect is the one most relevant to the use case argument for Bitcoin itself. The hashrate that protects the Bitcoin network is now being deployed at sites that are also serving frontier AI workloads. The economics of operating those sites are no longer dependent solely on Bitcoin's price cycle. They are partially subsidized by AI compute economics, which are growing for reasons that have nothing to do with cryptocurrency markets.

That cross-subsidy makes the Bitcoin network more resilient and the underlying asset more credible as long-duration collateral. A network secured by miners with diversified revenue is harder to disrupt than a network secured by miners dependent entirely on block rewards.

For an allocator, mining is now a backdoor into AI infrastructure with embedded Bitcoin exposure.

Family offices considering exposure to either Bitcoin or AI infrastructure are now evaluating overlapping assets. The public mining sector has effectively become a backdoor into the AI infrastructure build-out, with cash flows that derive from electricity arbitrage, compute leasing, and Bitcoin production. Each revenue line is independently valuable. The combination is more interesting than any single line.

The position can be expressed through public equities, private credit, or direct infrastructure investment. Each form of expression carries different risk and liquidity properties, but all three participate in the same underlying convergence. The investment case does not require any particular view on Bitcoin's price. It requires a view on whether AI compute demand and grid stability requirements will persist, both of which are well-supported by current evidence.

Next week: the sovereign and emerging-market channel, which operates on a different clock than every other use case covered so far.