The real bottleneck: power that can reach the site

AI data centers and Bitcoin mining operations compete for more than electricity at a low price. Both need substantial power capacity at suitable locations, where grid connections, substations, transformers, land, and permits can take years to secure. AI facilities also need cooling and network infrastructure capable of supporting high-density computing.

A low wholesale electricity price therefore does not guarantee that a project can get the power it needs. Local infrastructure and the time required to connect a facility can be the binding constraints.

The International Energy Agency (IEA) estimated that data centers consumed approximately 415 terawatt-hours (TWh) of electricity globally in 2024, about 1.5% of global electricity use. Its 2025 Base Case projected consumption of roughly 945 TWh by 2030. These figures cover data centers overall, not AI alone; AI is an important driver of growth. The IEA also notes that a data center can become operational in two to three years, while the energy infrastructure needed to serve it generally takes longer to plan and build (IEA, 2025).

That timing mismatch helps explain the appeal of existing powered sites, including some Bitcoin mining facilities. Where both uses are feasible, AI demand gives site owners another potential use for their available power capacity.

Why electricity costs matter to Bitcoin mining

Bitcoin mining uses application-specific integrated circuits (ASICs) to repeatedly hash candidate block headers, seeking a hash value less than or equal to the network's current target. On Bitcoin mainnet, the target is recalculated every 2,016 blocks—approximately every two weeks—to aim for an average block interval of ten minutes. A lower target corresponds to higher difficulty and more expected computational work per block (Bitcoin developer documentation).

In pooled mining, miners submit shares that meet an easier, pool-set target as evidence of contributed work. Their earnings depend on the pool's payout method, rather than solely on whether their own machine finds a network block (Bitcoin mining guide).

ASICs typically operate around the clock when economically justified, making electricity a major recurring operating cost. Mining economics also depend on fleet efficiency, commonly expressed in joules per terahash (J/TH), network difficulty, block subsidy and transaction fees, pool fees, uptime, and Bitcoin's market price. Bitcoin price changes the fiat value of mining earnings; it does not directly increase the amount of BTC a machine produces.

The decision to keep operating for another hour differs from the decision to invest in a mining business. In the short term, operators compare expected incremental revenue with avoidable operating costs, while considering power-contract obligations and any compensation or opportunity to sell power back. Long-term profitability must also account for hardware investment and other fixed costs. Accounting depreciation should not be treated as an electricity-like cash cost that disappears whenever a machine is switched off.

Why AI demand changes the value of a powered mining site

AI and high-performance computing (HPC) colocation can offer site owners contracted, dollar-denominated hosting revenue as an alternative to self-mining. That creates another business option, but it does not establish that AI hosting will always be more profitable.

The comparison involves more than projected revenue. Owners need to consider construction costs, financing, utility service terms, cooling, backup power, fiber connectivity, delivery timelines, and customer contract terms. A mining site's existing connection may offer a head start only if its capacity and service arrangements suit the proposed AI facility.

For an owner of suitable infrastructure, the options can include continuing to self-mine, hosting third-party miners, or developing AI/HPC colocation capacity. These activities carry different investment requirements and commercial risks.

Broader data-center demand provides context for this competition. Berkeley Lab's United States Data Center Energy Usage Report: 2025 Update, published in June 2026, projects 649 TWh of U.S. data-center electricity consumption in its 2030 Reference Case. Its broader scenario range corresponds to approximately 9.5%–15.3% of U.S. electricity use. These are modeled estimates for all data centers, not AI-only demand or forecasts for individual sites (Berkeley Lab, 2026).

Such projections help explain interest in suitable powered land, but they cannot establish the commercial viability of a particular conversion.

Mining infrastructure provides a head start

Existing land and electrical infrastructure can shorten development time. Mining containers and air-cooled mining buildings, however, are not automatically ready for dense GPU deployments. An AI hosting project may require substantial changes to cooling, electrical distribution, backup systems, and networking, as well as additional utility approvals.

Core Scientific's June 2024 CoreWeave agreements provide a historical example. The company described approximately 280 megawatts (MW) of site capacity intended to support about 200 MW dedicated to HPC hosting, with construction and site modifications required (Core Scientific, June 2024). By February 2025, it reported that its combined CoreWeave contracts covered approximately 590 MW of critical IT load across six sites (Core Scientific, February 2025).

Critical IT load describes power capacity for the IT equipment. It is different from total facility capacity, which also supports cooling and other systems, and from actual electricity consumption measured over time. The example illustrates the development of new hosting infrastructure at powered sites, rather than the reuse of Bitcoin ASICs for AI workloads.

Flexibility matters, but varies by facility

Bitcoin mining can often reduce electricity consumption quickly by switching off machines. Whether an operator chooses to do so depends on power prices, contract terms, and mining revenue conditions.

One relevant measure is hashprice: expected mining revenue per unit of hashrate per day, which can be expressed in BTC or a fiat currency such as USD. Higher USD-denominated hashprice can make continued operation attractive at electricity prices that would otherwise prompt curtailment.

In its May 2025 planning report, the Electric Reliability Council of Texas (ERCOT) treated cryptocurrency mining as Large Flexible Load and assumed other loads were firm for that planning exercise. This was a specific modeling distinction, not a universal statement about every mining or AI facility (ERCOT, May 2025). ERCOT's February 2025 resource-adequacy methodology also used mining economics to estimate price-responsive curtailment, illustrating why flexibility depends on operating conditions (ERCOT, February 2025).

AI facilities may face customer uptime commitments that constrain curtailment, although their flexibility varies with workloads, infrastructure, and contracts. Mining's ability to shut down readily can be valuable, but actual grid support depends on when and how that flexibility is used.

Conclusion

AI compute changes the opportunity cost of allocating suitable power capacity to Bitcoin mining. A powered site may support another business model, but converting it requires investment and depends on local infrastructure and customer demand.

For miners, electricity prices, hardware efficiency, and mining revenue remain central. For AI hosting, construction costs, service requirements, and customer contracts also shape the outcome. The more attractive use depends on the particular site and its economics; there is no fixed winner across all locations.

Frequently Asked Questions

What are AI compute and Bitcoin mining competing for?

They can compete for electricity and the infrastructure needed to deliver it at suitable sites, including grid connections, substation and transformer capacity, land, and permits. Local delivery constraints can matter as much as the electricity price.

Is AI hosting always more profitable than Bitcoin mining?

No. The comparison depends on power costs, construction and financing requirements, customer contracts, and mining economics. Higher projected hosting revenue alone does not establish higher profitability.

Can Bitcoin mining facilities easily become AI data centers?

Existing power infrastructure can help, but the site must meet the proposed AI facility's requirements. Cooling, electrical distribution, backup power, and networking may need substantial upgrades. Bitcoin mining ASICs cannot be repurposed to run AI workloads.

Does Bitcoin mining always support grid reliability?

No. Mining can reduce load quickly, but whether that supports the grid depends on the timing, extent, and terms of curtailment. Operators' decisions depend partly on power prices and mining revenue.

Why are electricity costs so important to miners?

Mining ASICs consume electricity while hashing, making power a major operating expense. Profitability depends on the relationship between mining revenue and costs, including electricity, equipment investment, and other expenses.

References

• IEA — Energy demand from AI (2025)
• Berkeley Lab — United States Data Center Energy Usage Report: 2025 Update (published June 2026)
• Bitcoin Developer Documentation — Block Chain
• Bitcoin Developer Documentation — Mining
• Core Scientific — CoreWeave hosting agreement presentation (June 2024)
• Core Scientific — CoreWeave contract expansion (February 26, 2025)
• ERCOT — Capacity, Demand and Reserves Report (May 2025, revised)
• ERCOT — Monthly Outlook for Resource Adequacy (February 2025)