Bernstein Sees Bitcoin Miners Becoming AI’s Electricity Landlords, With TeraWulf and Cipher Poised for Big Gains

Bitcoin’s infrastructure builders are increasingly monetizing something AI can’t scale without: immediate, large blocks of power. That’s the core of Bernstein’s new coverage, which frames miners as the gatekeepers of energized real estate for AI data centers and starts TeraWulf (WULF) and Cipher Digital (CIFR) at Outperform.

Here’s the structural shift that matters: miners spent a decade mastering site acquisition, interconnection, and low-cost electricity procurement. As generative AI’s “time-to-compute” premium tightens budgets and deadlines, those same campuses—already wired and ready—have become scarce inventory. Over the last two years, miners inked 17 agreements exceeding $110 billion, allocating roughly 6 gigawatts of capacity to hyperscalers like Google, Amazon, Microsoft, Nvidia, and CoreWeave. That footprint represents about 10% of AI data centers now being built in the U.S. The mining sector’s pipeline totals roughly 30 gigawatts of planned power, positioning it as an onramp for compute-hungry tenants.

Bernstein’s model implies a rapid re-rating of miner cash flows as AI colocation ramps. Across its coverage, AI revenue is projected to climb ninefold—from $1.2 billion in 2026 to $10.7 billion by 2030. Within that, TeraWulf—anchored by a partnership with Fluidstack and Google—is forecast to generate about $1.7 billion in AI revenue by 2030, with EBITDA margins approaching 84%. Cipher Digital, whose client mix skews heavily to hyperscalers, is projected at roughly $1.2 billion in AI revenue and near-93% margins.

The business model shift is deliberate: long-term, take-or-pay colocation contracts convert previously volatile mining earnings into more predictable infrastructure-like returns. Financing is following suit. Project lenders are covering roughly 75% to 85% of build costs, with debt coupons well below the returns embedded in the leases. That spread matters—particularly as miners blend self-mining with recurring AI lease income to reduce cyclical exposure.

My lens here is simple: the advantage isn’t just megawatts; it’s speed to energized square footage. AI models depreciate in usefulness fast if training or inference is delayed. Miners that can deliver pre-energized shells—power, cooling, and distribution in place—arbitrage that urgency. It changes their optionality: they can pivot between Bitcoin hashprice cycles and fixed-fee AI leases without idling stranded assets. Psychologically, that stability attracts a different investor base—more infrastructure and credit-savvy capital—potentially lowering the sector’s cost of capital and raising acceptable leverage.

There are trade-offs. Concentration risk shifts from Bitcoin price to counterparty and policy. Take-or-pay sounds bulletproof until you test it against tenant credit, technology obsolescence, or interconnection delays. Power markets are political by nature; community pushback, permitting timelines, and carbon scrutiny can compress returns if not proactively managed. Strategically, the best operators will over-communicate grid benefits (demand response, curtailed renewables offtake) and diversify power sources to reduce basis risk.

On selection, I’d prioritize: - Contract tenor and escalators, plus tenant diversification across hyperscalers and AI service providers. - Interconnection queue status and substation readiness; months matter more than ever. - Power mix and hedging—fixed-price PPAs vs merchant exposure, and regional grid reliability. - Balance sheet structure—match-funded project debt, covenants aligned with lease cash flows, and clear capex schedules.

For context, both WULF and CIFR shares are lower on the day, yet have ripped year-to-date in 2026—up roughly 122% and 69% respectively—suggesting the market is already leaning into the AI colocation thesis while still digesting execution risk. That’s consistent with a transition story: near-term volatility, medium-term rerating if delivery matches contracted ramps.

If compute demand keeps accelerating, control over energized capacity becomes as strategic as chips. Miners that consistently deliver power-on timelines, lock in bankable contracts, and maintain grid credibility could compound like infrastructure platforms rather than commodity producers. In a cycle where “faster to watts” beats “cheaper per watt,” the landlords of electricity will set the terms.