Energy, Not Edicts: Why U.S. Bitcoin Hashpower Is Ceding Ground to China and AI

America’s ambition to lead Bitcoin mining runs headfirst into a harder constraint: where surplus, flexible power actually exists. In 2025, North American mining pools steadily lost share of newly minted BTC even as Washington talked up dominance. The driver isn’t a lack of will; it’s the collision between AI’s insatiable compute buildout, miner economics at cycle lows, and a global energy map shifting underfoot.

Here’s the on-chain reality. By December, Foundry USA, MARA Pool, and Luxor Technologies collectively produced 35% of Bitcoin blocks, down from more than 40% in January. That erosion coincided with record-weak mining profitability when energy costs are included. JPMorgan pegs average daily revenue at just $38,700 per EH/s in December—down 32% year-over-year—while U.S. power prices marched higher over the past year. In that environment, operators often do the rational thing: redeploy scarce megawatts to higher-yield workloads.

AI has become the gravitational center. Industry leaders privately concede they have a fiduciary duty to test whether their power contracts, substations, and interconnections earn more serving high-performance computing than hashing. AI’s workload density and willingness to prepay for capacity can dwarf mining’s cash flows and compress the decision timeline. That’s why you’re seeing miners shift from “hashrate-at-all-costs” to “infrastructure-first” positioning.

Hut 8 is the cleanest example of the pivot. Once framed as a pure-play miner, the Miami-based firm is increasingly an energy infrastructure company. In December it announced a collaboration with Anthropic to develop U.S. data center capacity sized for AI-scale demand. In parallel, the Trump family moved from rhetoric to exposure: Eric and Donald Trump Jr. co-founded American Bitcoin last March; Hut 8 holds an 80% majority stake. Eric Trump recently toured the company’s Texas site, showcasing roughly 35,000 machines and claiming the facility mines “about 2%” of the world’s Bitcoin supply. The symbolism is clear; the economics are the constraint.

China underscores the other side of the equation. Despite a formal mining ban since 2021 and renewed scrutiny reported in December, hashpower has resurfaced—especially in Xinjiang. The region’s grid is dispersed, heavy on fossil generation, and far from Beijing’s center of enforcement, which invites risk-tolerant operators. More broadly, China has been rapidly expanding power generation capacity. If you treat Bitcoin mining as a buyer of last resort for stranded or off-peak energy, then more capacity—and more flexibility—naturally translates into more competitive bids for blocks. Activity across the Middle East and Russia adds to the rebalancing.

Publicly traded miners reflect this shift in posture. Hashrate expansion plans that looked like an arms race a few years ago have paused in many boardrooms. Some companies are reallocating power to high-performance computing rather than adding ASICs, preferring steadier contracted revenue over volatile block rewards. The psychology here is pragmatic: in a capex-heavy business, CFOs favor bankable offtakes, especially when ASIC breakevens creep up.

Upstream, the equipment market is feeling the whiplash. Bitmain, which has historically controlled an estimated 80% of global mining rig sales, has faced cooler demand and reportedly leaned into self-mining to monetize inventory—plugging hardware wherever capacity exists, from the U.S. to the Middle East and Central Asia. Scale back production too much, and you risk losing wafer allocations from TSMC later; keep building into oversupply, and margins compress. It’s a narrow corridor with little room for error.

Many in Washington prefer a story about policy will. The White House has floated ambitions to mine the remaining BTC domestically, yet rhetoric doesn’t change interconnection queues, regional power shortages, or AI’s willingness to pay a premium for electrons. If U.S. states continue prioritizing AI campuses—sometimes with bespoke incentives—Bitcoin miners will often find themselves price-takers, cycling rigs down during peak hours or exiting entirely. That isn’t anti-crypto; it’s grid triage.

The strategic takeaway is simple, if inconvenient: geography follows energy. Hash migrates to where kilowatt-hours are abundant, cheap, and politically tolerable—even if that means desert substations, coal-adjacent provinces, or countries comfortable with gray-market activity. In the U.S., miners that endure will likely look like energy operators first, compute allocators second, and Bitcoin specialists third. They will arbitrage day-ahead markets, co-locate with renewables, and treat AI and mining as interchangeable modular loads.

For market participants, watch three signals: local marginal prices and curtailment trends; AI capex and prepayment behavior; and enforcement cycles in places like Xinjiang that can swing illicit or quasi-tolerated capacity on and off. If those dials keep moving the way they have, North America’s block share may stabilize only when infrastructure expansion catches up—or when AI’s growth curve normalizes. Until then, the fulcrum isn’t who talks loudest; it’s who controls the next marginal megawatt.