Let’s face it, cryptocurrency markets are in decline. Between extreme altcoin dilution, disruptive geopolitical conflicts, and steady inflation, mining Bitcoin just isn’t as profitable as it used to be.
Halving events reduce block reward issuance by 50% every four years. With a deflationary supply, the value of each Bitcoin token issued as a block reward increases during halving’s, but the number of tokens issued decreases.
This means that although the price per token rises after halving’s because fewer tokens are being issued as block rewards, mining profitability has waned. Bitcoin mining companies still need to make a profit, especially to cover the cost of their expensive mining infrastructure, which can output massive computing power.
As mining profitability has declined, the computing power of mining setups is shifting from Bitcoin mining to AI inference.
Why are miners shifting to AI?
Training large AI models requires intensive computing power, typically supplied by massive data centers with high GPU density. Getting these data centers approved and permitted, and then actually constructing them, is time-consuming; if there’s anything that hyperscalers like AWS don’t have in the race towards Artificial General Intelligence (AGI), it’s time.
Bitcoin mining infrastructure can provide sufficient computing power to efficiently train models right now, essentially acting as AI data centers.
The profitability of AI compute provisioning is less volatile than Bitcoin mining, and global demand for AI compute is skyrocketing, providing a consistent revenue stream for miners.
How do Bitcoin miners fulfill the needs of AI?
Training AI systems requires coordinating GPUs with large memory and high throughput to process many computations in parallel across the same model. Additionally, training AI requires massive amounts of power.
Large-scale Bitcoin mines offer substantial electrical capacity and established infrastructure for industrial-scale power delivery, enabling them to run resource-intensive AI workloads continuously.
Additionally, AI development requires cooling and serious heat rejection capabilities. Bitcoin mines offer high-volume airflow, warehouse-style ventilation, and immersion cooling—in which GPU components are submerged in coolant solutions—to disperse heat density.
So, Bitcoin miners are switching to power AI development because GPU mining facilities can meet the compute needs of model training.
What Bitcoin mining companies are making the shift?
Some of the top publicly traded US BTC mining companies have already announced plans to pivot toward executing AI workloads rather than mining Bitcoin.
In November 2025, Bitfarms announced that the firm’s mining facility in Washington would be fully converted to power AI workloads. Bitfarms signed a $128 million contract with a US-based data center infrastructure provider to convert the facility and install state-of-the-art cooling systems, enabling it to meet the intensive requirements of AI development.
Additionally, this past summer, TeraWulf signed a 10-year AI hosting agreement with cloud provider Fluidstack to deliver critical IT load for AI workloads at its Lake Mariner data center in Upstate New York.
What are the risks for Bitcoin?
These pivots signify the shift from Bitcoin mining to AI inference. Although such a shift will be more profitable for miners in the long term, it could create vulnerabilities for the Bitcoin network.
As miners stop mining Bitcoin and transition to providing compute for AI workloads, fewer validators will secure the Bitcoin network, potentially negatively impacting network resilience and increasing centralization risks as hash power becomes more concentrated due to fewer miners operating.
Nonetheless, Bitcoin’s price volatility and waning block reward profitability are making it difficult for miners to earn consistent revenue.
Conclusion
Shifting to provisioning compute for AI workloads will enable miners to repurpose their infrastructure for model training and earn revenue from it. It is almost a perfect storm.
With the declining financial feasibility of BTC mining and the insanely high demand for AI compute, it is entirely natural that miners would shift to powering AI development, as they already possess the hardware to do so.