Bitcoin Miners Have a New Competitor for Their Electricity: AI

Points of Focus

• AI is turning Bitcoin miners’ power infrastructure into a premium asset as Bitcoin mining revenue per MW gives way to long-term AI data center contracts.
• Bitcoin miners are giving AI companies a shortcut to power, while Iren, Hut 8, Applied Digital, Cipher, and others are repurposing existing mining sites, substations, grid connections, and data centers to meet AI demand without starting from scratch.
• AI could reshape Bitcoin’s hashrate economics if miners increasingly redirect electricity and infrastructure toward AI.

Bitcoin miners spent years solving and perfecting one of the hardest problems in digital assets: how to turn enormous amounts of electricity into computing power.

Now, AI farms are eyeing the infrastructure Bitcoin (BTC) miners built.

AI companies are increasingly both competing and cooperating with Bitcoin miners over computing power. The trend boils down to whether or not AI economics will eventually make Bitcoin mining the lower-value use of scarce power infrastructure.

It already appears that the answer could be “yes.”

• Iren is “decommissioning Bitcoin mining hardware and reallocating power and data center capacity toward AI Cloud Services.”
• Cipher has stopped mining at one site to convert it for Amazon Web Services.
• Core Scientific is increasingly monetizing its power capacity through AI/HPC contracts.
• TeraWulf has signed a 20-year AI lease worth approximately $19 billion.
• Hut 8 has turned its power portfolio into a platform for gigawatt-scale AI campuses.

Bitcoin mining infrastructure gives AI companies a shortcut through one of the industry’s biggest bottlenecks — access to power.

AI is bidding against Bitcoin for the same megawatts

Bitcoin mining converts electricity into hashing power. Its economics depend on BTC’s price, network difficulty, global hashrate, machine efficiency, and electricity costs.

Electricity is consequently the industry’s dominant operating expense. Cambridge’s Digital Mining Industry Report estimates that it accounts for more than 80% of miners’ cash-based operating costs.

Hashrate Index’s September 2026 data illustrates the revenue side.

Depending on fleet efficiency, Bitcoin mining operations are currently earning:

• Under 14 J/TH fleets — $133 per MWh
• 14-19 J/TH fleets — $97 per MWh
• 19-25 J/TH — $73 per MWh
• 25-38 J/TH — $50 per MWh

AI infrastructure is increasingly producing a very different set of numbers.

Applied Digital’s 300 MW Delta Forge 1 campus is covered by a roughly 15-year lease worth about $7.5 billion. The company estimates $11 million-$13 million of capital expenditure per MW.

Hut 8’s two Beacon Point leases cover 704 MW of critical IT capacity and represent $19.6 billion in contract value. Its broader AI portfolio totals 949 MW and $26.6 billion of base-term value.

Bitcoin figures describe mining revenue from electricity consumption, while AI figures are generally long-term lease or colocation revenue and require much greater capital investment, but the economic direction increasingly shows that a megawatt locked into a long-term AI contract may be worth considerably more than a megawatt pointed at the Bitcoin network.

Iren: The Bitcoin-to-AI transition in real time

Iren provides the clearest example of a Bitcoin-to-AI transition. As of June 30, 2026, Iren had approximately 5 GW of grid connection agreements and equivalents, around 40 MW of operating AI Cloud capacity, and approximately 380 MW of data center capacity supporting Bitcoin mining. Its installed mining capacity was about 23.2 exahashes per second (EH/s).

But Iren began decommissioning Bitcoin mining hardware and reallocating power and data center capacity toward AI cloud services, targeting substantial completion of the transition by the end of 2026.

Iren says recent three-year AI contracts generate more than $20 million of revenue per MW of IT capacity, with discussions around approximately $25 million per MW. It has also highlighted customer prepayments covering 45%-55% of GPU capital expenditure.

For Iren, AI capacity can generate several times the annual gross revenue of even highly efficient Bitcoin mining capacity, and the transition doesn’t have to start from zero. It already has land, power connections, substations, and data center infrastructure.

In essence, the Bitcoin mining business effectively paid to develop the first layer of the AI business.

AI vs. Bitcoin mining

A miner has exposure to BTC’s price, network difficulty, hashrate, and hardware cycles, all of which vary, and many of which are intentionally designed to get harder.

An AI infrastructure company can instead sign a 15- or 20-year contract and finance its infrastructure against predictable customer payments.

That’s a fundamental difference in revenue quality and predictability, at a time when Bitcoin’s price is still recovering after falling 50% in a year.

Hyperscalers and AI companies have enormous demand for computing capacity today. They cannot necessarily wait years for a new campus to acquire land, secure an interconnection, complete permitting, build substations, and finally become operational.

But Bitcoin miners have already spent years doing precisely those things, gaining infrastructure that includes:

• Industrial-scale grid connections
• Substations and switchyards
• Power contracts
• Land
• Data center shells
• Cooling systems
• Fiber connections
• Permitting expertise
• Utility relationships.

The Bitcoin aftermath

One thing AI companies don’t need? The ASIC. So, what happens to Bitcoin’s hashrate if everyone moves to AI?

If enough miners make the jump, hashrate growth slows. Increasing competition from AI workloads could contribute to a flattening of Bitcoin’s hashrate as miners redirect energy and infrastructure toward potentially more profitable uses.

If the hashrate falls, difficulty eventually declines, improving the economics for the miners that remain. So, while AI could reduce the marginal amount of electricity flowing into Bitcoin mining and slow the growth of network hashrate, the remaining Bitcoin miners could actually benefit from less competition.

The AI future

AI infrastructure is extraordinarily capital intensive. GPUs, networking, cooling, and electrical systems can require millions of dollars per MW, and today’s AI economics (with millions of dollars in yearly revenue) are not guaranteed forever.

If data center supply catches up with demand, GPU rental rates and infrastructure margins could fall. Bitcoin also retains an advantage that AI contracts do not: Its upside is uncapped. A major BTC price increase can dramatically improve mining economics.

Right now, Bitcoin companies are increasingly also becoming AI companies. Will the current energy economies hold? Where will the next megawatt go — toward the AI boom or toward a resurgent Bitcoin?

The industry’s future hashrate could increasingly be determined not just by the price of Bitcoin or the efficiency of mining hardware, but by one simple question:

How much is AI willing to pay for the megawatts that Bitcoin miners used to have to themselves?