Bitcoin Miners Migrate to AI: Profitability at Stake

What Happened?

Over the past weekend, Bitcoin mining difficulty dropped by 10%, according to Gizmodo. This is the largest percentage decline since July 2021, when China's mining ban caused a similar drop. The reduction reflects that a significant number of miners have disconnected their equipment, mainly due to falling profitability after the 2024 halving, which cut the block reward from 6.25 to 3.125 bitcoins. Added to this is the rise in global energy costs, which has further squeezed margins. Instead of continuing to mine, companies like Marathon Digital and Riot Platforms are repurposing their data centers to offer artificial intelligence (AI) services. Marathon Digital, for example, announced in January 2025 the conversion of one of its Texas facilities to house GPUs for AI model training, according to an official statement. Riot Platforms, meanwhile, has allocated 200 MW of capacity at its Rockdale, Texas farm for high-performance cloud computing services.

Why Is This Important?

This transition is not just a change of business; it represents a convergence between two compute-intensive industries. Miners possess electrical infrastructure, cooling systems, and hardware management that are directly applicable to AI model training. By migrating, they relieve pressure on the Bitcoin network — explaining the difficulty drop — and simultaneously inject processing capacity into the AI market, which faces a chronic GPU shortage. According to a March 2025 Bernstein report, GPU demand for AI exceeds supply by 30%, and traditional data centers cannot scale fast enough. Miners, with long-term electricity contracts and access to high-end hardware, can fill that gap more quickly and efficiently. Moreover, reusing existing infrastructure reduces the environmental impact of building new data centers, although AI also consumes significant energy. A University of Cambridge study estimates that Bitcoin mining consumes about 150 TWh per year, while AI could require an additional 85 to 134 TWh by 2027. The convergence could optimize global energy use.

Consequences for the Ecosystem

• For Bitcoin: Lower hash power, but greater efficiency for remaining miners. The network becomes more decentralized if small miners survive, though the difficulty drop also reduces network security in the short term. Historically, after similar events (like the 2018 crash), the network has recovered, but the current migration is more structural.
• For AI: More compute supply at competitive prices, which could accelerate the development of smaller, more accessible models. Companies like CoreWeave, which started as a crypto miner, have already become key AI infrastructure providers, with contracts from Microsoft and Google, according to The Wall Street Journal.
• For Investors: Shares of miners diversified into AI could appreciate. For example, Marathon Digital's stock rose 15% in the past month after announcing its AI pivot, while pure-play miners like Hut 8 fell 8%. JPMorgan analysts advise caution, as the transition requires significant GPU investments (up to $500 million per farm) and not all companies will succeed.
• For the Environment: AI also consumes a lot of energy, but reusing existing infrastructure avoids building from scratch. However, the carbon footprint depends on the energy source. In Texas, where many miners operate on renewable energy, the transition could be positive. In coal-dependent regions, the impact could be negative.

What Readers Should Know

This is not the first time miners have shifted activities: after the 2018 crash, many turned to cloud computing. However, the current scale is larger. In 2018, difficulty dropped 15% in a month, but the migration was temporary. Now, the trend seems more permanent due to structurally lower post-halving mining profitability. Experts like Alex de Vries, founder of Digiconomist, note that this trend could last as long as mining profitability remains low. However, if Bitcoin recovers (e.g., surpassing $100,000), some might return. The key is data center flexibility, which can switch between mining and AI based on prices. Companies like Hive Blockchain have already developed software that allows switching between workloads in minutes. Additionally, the AI market offers long-term contracts (1 to 3 years) with fixed prices, reducing revenue volatility. In contrast, mining depends on Bitcoin's price and difficulty, which are unpredictable.

"Bitcoin miners are doing what any business would do: seeking the highest return for their infrastructure. AI offers stable contracts, while mining is volatile," commented an industry analyst to CoinDesk.

In summary, we are witnessing a global reallocation of computational resources that will benefit AI in the short term, but also leaves Bitcoin more dependent on long-term committed miners. The Bitcoin network could become more efficient, but with less redundancy. For investors, diversification into AI is a bet on the future, but with risks: AI competition is fierce, and margins could shrink as more miners enter the market. Ultimately, this convergence could redefine the technological landscape, where crypto mining infrastructure becomes the backbone of high-performance computing.