Will miners sell or pivot amid AI power land‑grab?

Bitcoin network difficulty reached 136.04 trillion on Sept. 4, while dollar hashprice slipped to about $52 per petahash per day this week. Per Hashrate Index, the last adjustment set a new high for difficulty, and the forward market now prices an average hashprice near $49.17 per PH per day for the next six months.

The squeeze leaves miners deciding whether to sell inventories, consolidate operations, or pursue high-performance computing revenue tied to artificial intelligence.

The production backdrop is firm. The seven-day average hashrate sits near one zettahash per second, while transaction fees contribute a little over 1% of block rewards on recent averages.

That mix compresses gross margins at the same time retail power prices and wholesale data center rents trend higher. Global colocation pricing averaged $217.30 per kilowatt per month in the first quarter, with tight supply in major hubs, per CBRE’s Global Data Center Trends 2025.

Strategic optionality is widening as compute demand reorders the power stack.

CoreWeave agreed to acquire Core Scientific earlier this year in an all-stock transaction that implies roughly $9 billion of equity value. The acquisition would consolidate about 1.3 gigawatts of installed capacity with more expansion potential.

In its deal materials, the buyer outlined lease efficiency gains and operating synergies by 2027, while the transaction is part of the broader AI buildout competing for grid access across North America. The direction of travel is clear: AI workloads are now a core alternative for power and land that previously skewed toward proof of work.

Public market signaling has also shifted with the debut of American Bitcoin Corp. The company began trading on Nasdaq as ABTC after completing a merger with Gryphon Digital Mining. Corporate filings detail a controlled structure after the combination, with former American Bitcoin holders owning about 98% of the combined company on a fully diluted basis.

The model emphasizes accumulation alongside self-mining, creating another lever for treasury strategies that may dampen or amplify market sales depending on spreads between mining cost, spot price, and financing terms.

Power constraints and policy continue to set near-term supply behavior.

In Texas, miners commonly curtail during the Four Coincident Peak season to manage costs and capture credits, a pattern reflected in Riot Platforms’ June operating update. Curtailments can lift hashprice temporarily and shift revenue timing, but they also illustrate why forward hedging has become standard. Luxor’s market shows an actively traded curve with mid-market quotes published on the Hashrate Forward Curve.

Against this backdrop, break-even math is simple but unforgiving. Using representative efficiency bands and current economics, the ranges below illustrate approximate breakeven power prices, expressed in cents per kilowatt hour, at a $53 per PH per day hashprice and nominal pool fees.

The inputs reference published specifications for the Antminer S21 and WhatsMiner M60S, along with incremental firmware gains evidenced by LuxOS testing.

These thresholds imply that fleets paying above single-digit power rates will feel pressure if hashprice tracks the forward average. That pushes treasurers toward hedges on the hashrate curve, deeper curtailment during high-priced hours, and non-mining revenue.

The last category includes AI colocation and managed GPU services, where contracted rents are quoted per megawatt per year and often load follows compute.

Recent contracts frame the revenue step change.

TeraWulf disclosed more than $3.7 billion of expected hosting revenue under multi-year agreements, with public reporting estimating an annualized take rate near $1.85 million per megawatt on the initial tranche.

The comparison below uses those public figures and CBRE’s rent benchmarks to show the order of magnitude gap between mature AI colocation and current mining cash generation per power unit at prevailing hashprice.

The delta does not automatically mean every miner should pivot.

Retrofits require capex, liquid cooling, and higher-density racks, which can saturate existing transformers, and contractual take-or-pay obligations can limit near-term flexibility.

Still, the combination of tight colocation supply and announced consolidation, such as CoreWeave’s deal, will likely keep AI rents firm through year-end, which factors into treasury choices whenever bitcoin’s fee share remains low.

Miners able to monetize demand response programs, like the ERCOT 4CP framework, and tune fleets with efficiency firmware can widen their breakeven bands without selling coins.

Case studies illustrate the choice set. Iris Energy continues to expand GPU capacity and cloud revenue alongside self-mining, using a dual track that stabilizes cash flows against hashprice volatility. 

American Bitcoin presents a treasury-led approach combining on-balance sheet accumulation with mining, with control details and share counts in the SEC filing. Those paths sit alongside pure play hosting that captures AI demand and infrastructure premiums.

The near-term market question is whether balance sheets become a supply source by year-end. If hashprice follows the forward curve and fees remain near current prints, miners above the single-digit cost bands are more likely to raise cash by selling coins or locking in forward sales of hashrate.

If AI colocation ramps up on previously announced contracts, some of that selling could be offset by compute reallocation and hedges already layered in at summer premiums.

The balance of those forces will determine how much miner supply reaches exchanges during the fourth quarter.

Mentioned in this article