Seventy percent of leading Bitcoin miners are utilizing AI-generated revenue to navigate the bear market.

Seven out of the top ten miners by hashrate are currently engaged in AI or high-performance computing projects that are already producing revenue, while the remaining three are preparing to implement similar strategies.

This transition aligns miners’ active land and connections with guaranteed income from GPU clients, establishing an additional revenue stream that competes with operating ASICs at maximum capacity.

AI collaborations transform mining economics and investor interest

TeraWulf established a benchmark by entering into two 10-year hosting agreements with Fluidstack, amounting to around 200 MW at Lake Mariner.

As reported by Barron’s, Google is supporting a segment of Fluidstack’s lease obligations, up to approximately $1.8 billion, and has obtained warrants that could represent about 8 percent of TeraWulf. The financial details of the deal suggest approximately $1.85 million per MW annually in headline revenue over the contract duration, which many miners now consider a standard when attracting AI clients.

Core Scientific has expanded its 12-year partnership with CoreWeave to include about 70 MW of additional HPC capacity, with operations expected to commence in the latter half of 2025. Bitdeer continues to run a commercial AI cloud utilizing NVIDIA DGX systems, while Iris Energy reports an AI cloud operation based on H100 and H200 GPUs.

Others are developing the infrastructure for the upcoming wave. CleanSpark announced on October 29 that it has acquired 271 acres and approximately 285 MW of long-term power in Texas for what it describes as a next-generation AI and HPC campus. In August, Marathon agreed to purchase 64 percent of Exaion, a subsidiary of EDF, to enhance its global AI and HPC capabilities, with an option to increase its stake to 75 percent by 2027.

Riot is evaluating the conversion of around 600 MW at Corsicana for AI or HPC and has halted part of its mining expansion, leading to a reduction in its year-end 2025 hashrate forecast from 46.7 EH/s to 38.4 EH/s. Bitfarms has engaged consultants to perform a feasibility study and has been promoting its sites to AI clients.

Cipher Mining is reported to have a multi-year agreement with Fluidstack linked to Google, although not all terms are detailed in a single primary filing. Abu Dhabi’s Phoenix Group has indicated plans to expand data-center capacity beyond 1 GW, focusing on AI, and is considering a U.S. listing to finance this growth.

Bitcoin miners by hashrate and AI engagement

The financial rationale hinges on power and predictability.

In the current network context of approximately 1.08 to 1.10 ZH/s and 144 blocks per day, with fees fluctuating between roughly 0.3 to 2.0 BTC per block, one MW of modern ASICs operating at about 17 J/TH equates to roughly 0.059 EH/s of hashrate.

This portion of the network generates approximately $1.0 to $1.6 million per MW annually in gross mining revenue before accounting for power and operational expenses, at a bitcoin price nearing $104,000, based on CoinWarz data for price and hashrate. The midpoint of this range, around $1.2 to $1.3 million, falls short of the $1.85 million per MW per year suggested by TeraWulf’s AI agreements.

Power costs, capital expenditures (capex), and utilization influence margins in both models. Nevertheless, the contracted nature of AI hosting has emerged as a significant aspect for equity investors seeking more stable cash flows rather than mere exposure to risk and fees.

The macro demand for data-center power provides the context. McKinsey reports indicate that U.S. data-center electricity consumption could reach approximately 606 TWh by 2030 as AI workloads increase. ERCOT anticipates record peak demand over the next five years, with data centers being a major contributor, as analyses predict around 35 GW of peak data-center load by 2035.

Utilities are adapting, with American Electric Power raising its five-year capital plan to $72 billion as it navigates a pipeline of customer-backed contracts and over 190 GW of load requests in development, according to Reuters. These figures align with miners’ assertions that their grid connections, substations, and land holdings are now valuable resources for AI campuses, not just for exahash.

This realignment alters what is significant within the mining league table.

A miner that allocates new megawatts to AI may experience slower headline hashrate growth compared to a dedicated operation. However, its enterprise value can increase through contracted revenue, power flexibility, and longer-term agreements.

Core Scientific’s expansions with CoreWeave solidify a 12-year commitment to this model. CleanSpark’s 285 MW initiative and Marathon’s acquisition of Exaion encourage miners to develop and manage mixed-use campuses where GPUs, miners, and sometimes standard colocation can share infrastructure. Riot’s public assessment of 600 MW at Corsicana illustrates how swiftly the mix can shift when a site already possesses transformers, switchgear, water rights, and fiber infrastructure.

There are limitations. ERCOT interconnection timelines, gas turbine availability for new peakers, and transformer lead times all influence how quickly high-density facilities can be activated. GPU supply remains a variable as Blackwell and successor components ramp up and as hyperscalers allocate inventory to internal projects.

On the crypto front, any alteration in fee structures that significantly raises fees per block could narrow some of the per-MW revenue gap between mining and AI hosting. An increase of about 0.5 BTC per block in sustained average fees translates to roughly $0.2 to $0.3 million per MW annually in miner gross revenue at current price levels, based on the straightforward share-of-network calculations mentioned earlier.

Investors are focusing on revenue composition, rather than solely on exahash.

Contracted AI megawatts and revenue per MW per year are becoming the new metrics to monitor. The $1.5 to $2.0 million per MW per year range is emerging as a practical standard for high-density hosting in the U.S., with TeraWulf’s disclosed figure serving as a current benchmark.

Utility capex plans and interconnection queue updates are now as pertinent to miner forecasts as ASIC delivery timelines. As U.S. spot power tightens, miners with already energized land, permitted sites, and available substations can capitalize on that flexibility more rapidly than new entrants.

The international dimension adds significance. Marathon’s collaboration with Exaion connects a U.S. miner to an EDF affiliate within the French power system, aligning GPU hosting with state-adjacent energy resources.

Phoenix Group’s strategy to expand in the Gulf, while considering a U.S. listing, incorporates sovereign power economics into the equation for AI infrastructure.

Such arrangements could draw more miners into partnerships where utilities or energy investors anchor long-term contracts in exchange for capacity rights, priority interconnections, or equity stakes.

For crypto fundamentals, this shift could decelerate the rate at which network hashrate grows through 2026 if significant portions of new power are directed toward GPUs instead of ASICs. The network will still increase hash as new sites become operational and as older fleets are updated, yet the growth rate may flatten compared to the previous surge.

This would not prevent capital from flowing into mining, as high bitcoin prices and fee increases can still enhance returns; however, it renders the hashrate leaderboard a less reliable indicator of equity value than it was in earlier cycles.

Below is a brief overview of the current status of the largest listed miners. The status indicates whether AI/HPC is already generating revenue or is still in the planning or evaluation stages, based on company disclosures and mainstream reporting.

Bitcoin Miner	Hashrate (EH/s)	% of Global Network	AI/HPC Involvement	Status
Marathon Digital Holdings	57.4	5.3%	Acquiring 64% of EDF’s Exaion to expand AI/HPC infrastructure	Revenue
CleanSpark	50.0	4.6%	Building 285 MW AI/HPC data-centre campus in Texas (contracts under development)	Revenue
Iris Energy (IREN)	45.4	4.2%	Operating renewable-powered GPU AI cloud clusters with H100/H200 systems	Revenue
Riot Platforms	36.5	3.4%	Evaluating AI/HPC repurpose of 600 MW Corsicana facility (paused mining expansion)	Planning
Bitdeer Technologies	35.0	3.2%	Running commercial AI cloud service using NVIDIA DGX H100/H200 GPUs	Revenue
Cipher Mining	23.6	2.2%	Reported multi-year AI data-centre leases (AWS & Fluidstack, ~$8.5 B total)	Revenue
Core Scientific	19.1	1.8%	Hosting AI/ML workloads for CoreWeave under 12-year contract (~70 MW)	Revenue
Bitfarms	19.5	1.8%	Conducting HPC/AI conversion feasibility with Appleby Strategy Group	Planning
TeraWulf	12.8	1.2%	Signed 10-year AI hosting contracts (> 200 MW, Google-backed Fluidstack)	Revenue
Phoenix Group*	15.0 *	1.9% *	Expanding toward 1 GW hybrid data-centre capacity for AI/HPC by 2027 (planned)	Planning

What to monitor now is straightforward and quantifiable. Keep track of contracted AI megawatts and revenue per MW per year in new filings, utility capex trends, and ERCOT load updates, along with thirty-day averages for bitcoin fees relative to the subsidy using sources like CoinWarz.

These data points will indicate how much mining power transitions to GPUs, how swiftly campuses become operational, and how the per-MW revenue gap changes. The largest miners are already implementing that strategy.

The post 70% of top Bitcoin miners are already using AI income to survive bear market appeared first on CryptoSlate.