For over a decade, Bitcoin mining farms have been heavily criticized in the energy and technology sectors. Their massive electricity consumption has sparked congressional hearings, ESG rating downgrades, and persistent public criticism. Yet today, these farms have secured 15-year leases with companies like Microsoft, Google, and Anthropic. The farms themselves have remained virtually unchanged. In fact, if there's anything in common among these farms over the past decade, it's the crisis itself. So, what happened? There's an interesting adage about crises: "The best opportunities often come from the worst crises." This perfectly describes the experience of Bitcoin miners. From July 2016 to April 2024, they experienced three halvings. Each halving halves the block reward, forcing miners to search for cheaper electricity in increasingly remote corners of the US power grid, including rural West Texas, Georgia, and the North Dakota plains. The weak were eliminated. Some companies adapted in time. Others learned their lesson later. In today's story, I'll explain how the surge in investment in AI infrastructure aligns with miners' increasing computing power and processing capacity, helping them to find new life. Now let's continue. Halving—The First Turning Point The first test of survival for Bitcoin miners occurred in April 2024, the most recent Bitcoin halving event. Every halving is a stress test. But with each halving, the reward is halved, and the challenge doubles. The April 2024 Bitcoin halving reduced the block reward from 6.25 BTC to 3.125 BTC. In the week following the most recent halving, the price of hashrate fell from $0.12 per terahash to $0.047. Hashrate refers to the revenue a miner expects to earn per unit of computing power. By the first quarter of 2026, the price of hashrate had fallen to a five-year low of $0.023 per terahash per day. Currently, the average cost of producing one Bitcoin is approximately $81,000. If other non-productive costs required for miners to maintain operations are also included, the total cost of mining each Bitcoin far exceeds $115,000. Bitcoin is currently trading at $70,760. Its price has never exceeded $80,000 in the past three months. Do the math yourself. The Bitcoin mining industry can only continuously pursue lower mining costs, and it has no control over the price of Bitcoin. Miners whose main income came from the price difference between mining Bitcoin and selling it on the open market suddenly found themselves with losses on their financial statements. So, they switched to holding the mined Bitcoin. Their idea was to wait for the price of Bitcoin to rise enough to generate a positive return. This strategy worked until the price of Bitcoin rose. But market fluctuations are cyclical. Every bull market is followed by a bear market and a correction. The cryptocurrency market is no exception. 10/10 - The Second Turning Point October 10, 2025: A terrifying day for the cryptocurrency industry, witnessing the largest cryptocurrency liquidation in history. Since then, cryptocurrency prices have experienced a record drop, marking the beginning of a bear market. This led to the complete collapse of miners' "mine and hold" strategy. Some companies hesitated to change their strategies. However, others announced strategic transformations within 24 hours of the liquidation events. On October 11, Bernstein released a report redefining the role of Bitcoin miners, no longer viewing them as producers of computing power, but rather as holders of gigawatt-level secure grid access. Analysts called these miners "a key link in the artificial intelligence value chain." They unanimously agreed that IREN (formerly Iris Energy) was the top choice for successfully transitioning from Bitcoin mining to a cloud infrastructure provider focused on artificial intelligence. Galaxy Digital, a leading digital asset and AI infrastructure provider, announced it has raised $460 million to convert its Helios mining farm in Texas into a CoreWeave high-performance computing (HPC) campus on a 15-year lease, with projected annual revenue exceeding $1 billion. Following the 10/10 event, a series of systemic balance sheet liquidations followed, those that had defined the industry with a "mine and hold" strategy. Miners had spent at least 18 months accumulating Bitcoin as a reserve asset, viewing unsold Bitcoin as a sign of confidence. This stance began to waver under the pressure of the bear market, with Bitcoin's price falling approximately 40% from its all-time high of around $126,000 in 45 days. Some publicly traded miners who had never previously sold Bitcoin also began to sell. Marathon Digital (MARA), the third-largest publicly traded Bitcoin holder in the U.S., broke its record for continuous Bitcoin holdings by selling 15,133 bitcoins in three weeks. The company's CEO has consistently championed and drawn inspiration from this strategy, making it the largest corporate Bitcoin reserve. Less than two years ago, MARA's CEO and Chairman, Fred Thiel, announced that Bitcoin would become its strategic reserve asset. Just last month, Fred reversed course, acknowledging that selling Bitcoin "enhanced financial flexibility and increased strategic options as we expanded our business from pure Bitcoin mining to digital energy and AI/high-performance computing infrastructure." However, I wouldn't blame him. Tough times call for tough decisions. And MARA isn't the only one abandoning Bitcoin as a permanent strategic asset. While some investors increased their Bitcoin reserves following the liquidation events, others slowed their Bitcoin accumulation or publicly stated they no longer considered Bitcoin a strategic reserve asset. Bitfarms' CEO frankly admitted, "We are no longer a Bitcoin company." Ben Gagnon added that Bitfarms will focus on "building the infrastructure for the future of computing." CleanSpark took a different approach, treating its more than 13,000 Bitcoins as productive capital and allocating them to multi-tiered covered call options. Even if Bitcoin doesn't disappear from their balance sheets, they view it as a resource to strategically drive their infrastructure transformation. What seems like a misfortune may actually be a blessing in disguise. Transforming Bitcoin mining farms into artificial intelligence infrastructure is no easy feat. The cost per megawatt ranges from $8 million to $11 million, including new liquid cooling systems, triple-level power redundancy, high-bandwidth fiber optics, and network upgrades required for GPU training clusters. However, mining infrastructure, including cooling, power, and computing power, is closer than any other industry to meeting the needs of the AI and data center industries. Bernstein analysts noted in a report that miners' existing infrastructure could reduce deployment time by up to 75%. This view isn't shared by analysts alone. Deals made by these mining companies in recent months corroborate this.
IREN signed a $9.7 billion contract with Microsoft to provide GPU cloud hosting services at its Childries, Texas campus, making it the largest single deal to date between a miner and a hyperscale data center. Hut 8 entered into a $7 billion deal with Google-backed Fluidstack and Anthropic. Cipher Mining signed an $8.5 billion contract with AWS and Fluidstack. By the fourth quarter of 2025, Core Scientific's revenue from AI hosting (i.e., renting space in data centers to house IT equipment) will rise from 9% four quarters ago to 39%.
Surprise Moat
But why would hyperscale data center operators pay mining companies for data center space?
Time is of the essence.
To survive each halving of electricity prices, miners have had to chase cheaper power. To survive, they've had to take various measures: negotiating long-term power supply agreements, acquiring industrial land in low-cost energy corridors, building dedicated substations, and ensuring direct grid interconnection. Modern mining farms are equipped with dedicated high-voltage transformers, redundant power supplies, and thermal management systems designed to operate at full capacity 24/7. Perhaps this wasn't planned in advance; you might say the miners were just lucky. But who can be lucky enough to strike it rich while struggling to survive? Currently, public mining companies have approximately 6.3 gigawatts of operational capacity, with another 2.5 gigawatts under construction. In the United States, data center interconnection queues in most markets are 5 to 7 years. Microsoft's internal forecasts indicate that its data center resource shortage will continue until 2026 and beyond. This is why hyperscale data center operators overlook the lack of expertise in AI infrastructure among mining companies. Instead, they pay for substations, land use permits, utility relationships, and grid connections—features that often take years to materialize elsewhere. Mining companies can gradually improve their performance by adapting existing equipment for AI applications. MARA recently announced a $1.5 billion acquisition of energy infrastructure, bringing its total generating capacity to over 2.2 gigawatts. This allows MARA to transform a depreciating facility into AI infrastructure at a cost unmatched by other AI infrastructure builders. CEO Fred Thiel stated that these assets are off-the-shelf infrastructure that would have taken up to 10 years and cost $2 billion to $3 billion to build independently. He stated, "Electricity is a scarce input in the field of artificial intelligence, and with the planned acquisition of Longridge Energy, we will control a highly efficient, contracted energy platform." A Closed Window There's a pitfall in this story. Every megawatt of energy shifted from Bitcoin mining to AI infrastructure subsidizes the economic interests of those still mining Bitcoin. This lowers mining difficulty, making it cheaper for Bitcoin miners to mine a block. Perhaps some will still choose to use some equipment for Bitcoin mining in case prices drop. But this is limited to those who can afford to replace equipment or reserve mining equipment costs. Not everyone can do that. The reason is that those who use mining equipment for AI infrastructure cannot repeatedly switch between mining and AI. Mining is an interruptible process. When electricity is high, you can shut down your mining rigs. But AI and high-performance computing cannot do that. Once you lease or commit to using your computing power, you can't temporarily cancel the agreement and switch to mining Bitcoin with that equipment. However, for most miners, this isn't a viable option. They only have a very short window to complete the shift change, and such good fortune doesn't come often. Everything went so smoothly it was almost unbelievable. The Bitcoin halving squeezed the economics of mining to the limit. The subsequent 10/10 liquidation forced miners to face reality: holding Bitcoin during a bear market cycle wasn't a viable strategy. But the booming development of AI infrastructure came at the right time, giving miners both the incentive to transform and the assets needed to execute that transformation. This situation is unlikely to repeat itself. Mining companies that sign contracts today will reap the rewards of the next decade's economic benefits, while those who come later will miss out.
Weatherly
Catherine
Kikyo
Joy
Anais