Summary of Key Points
Bitcoin mining companies such as TeraWulf, Hut8, and Core Scientific are facing increasing difficulties in their mining operations (volatile coin prices, halvings that cut revenue in half, and the costly competition for computing power). As a result, they have converted their existing mines into AI data centers, renting out space and electricity to AI companies. The mines' readily available electricity, land, and grid connection capabilities perfectly meet the urgent need for computing infrastructure from AI firms, leading to a perfect match between the two parties. Some mining companies have signed long-term leases worth billions of dollars, which has significantly boosted their stock prices. However, this transition also comes with financial pressures, construction risks, and uncertainties regarding changes in AI demand. Essentially, mining companies have shifted from betting on Bitcoin prices to gambling on the future needs of the AI industry.
I. Mining Is No Longer Lucrative: The Industry's Weakness Revealed by Corporate Bankruptcies
The good times for Bitcoin mining companies were fueled by a bull market; between 2020 and 2021, the price of Bitcoin soared from $10,000 to $69,000, prompting companies like Core Scientific to invest heavily in mining equipment and build new mines (making it the largest miner in North America). However, this business model has three critical flaws:
1. High Losses When Prices Fall: In 2022, when Bitcoin prices plummeted, the costs of electricity and equipment remained unchanged, leading Core Scientific to run out of cash and go bankrupt. Another company, Riot, produced 46% more coins but saw its revenue drop by 15%, resulting in a net loss of $500 million.
2. Unstoppable Competition for Computing Power: Mining is like an arms race, with companies constantly upgrading to faster mining machines. The difficulty of the mining process increases automatically, causing older machines to consume more power and produce less output, necessitating constant replacements and significant financial expenditures.
3. Halvings Cut Revenue in Half: Every four years, Bitcoin's reward system is halved, significantly reducing mining profits. After the next halving in 2024, mining companies will earn half as much for the same amount of computing power, despite unchanged costs.
These issues have combined to turn mining from a profitable venture into a cash-straining endeavor, forcing companies to seek new alternatives.
II. A Perfect Match: AI Companies Need Electricity, and Mines Have It
Mining facilities possess an underutilized asset: large amounts of electricity and grid connection capabilities. Mining requires substantial power (for example, TeraWulf's mines use electricity from nuclear power plants), so these companies have already invested in the necessary infrastructure for connecting to the电网. AI companies, on the other hand, face significant challenges in obtaining sufficient power for training large-scale GPU models. Data center construction can take years, and even purchased GPUs may remain unused. For instance, CoreWeave, an AI cloud computing company that formerly mined Ethereum, struggled with storage for its GPUs and decided to rent a mine from Core Scientific. The collaboration between the two companies highlights the value of these facilities in the AI era.
III. The Transition Process: Selling Mining Equipment and Dividing Business Units
The transition of mining companies is not just theoretical; it involves concrete actions:
1. Raising Funds by Selling Mines: TeraWulf sold 25% of its shares in its nuclear-powered mine and used the proceeds to invest in an AI data center.
2. Separating Mining Operations: Hut8 spun off its mining equipment and operations into a separate subsidiary, while the parent company focused on renting out its AI data centers.
3. Direct Conversions: Core Scientific transformed its restructured mines into AI data centers and rented them to CoreWeave, shifting from being a major miner to an AI infrastructure provider.
The core of these changes is getting rid of outdated mining equipment and retaining the valuable assets of the mines.
IV. Benefits and Risks of the Transition
The benefits of this transition are clear:
- Stable Revenue: AI leases are typically long-term (15–20 years), providing a more stable income source compared to the volatility of Bitcoin prices. For example, TeraWulf's $19 billion lease led to a 19% increase in its stock price, and Hut8's $9.8 billion lease resulted in a 17% rise.
- Easier Financing: Long-term leases enable companies to secure loans from banks for data center construction.
However, the risks are also significant:
- Delayed Revenue: The benefits of these contracts are spread over a long period (20 years), meaning funds are not received immediately. TeraWulf's project will be completed in 2027, and Hut8's in 2028, with substantial upfront investments required for infrastructure.
- Construction Risks: Delays, rising equipment costs, and increased financing expenses can reduce profits. Additionally, if AI demand changes (e.g., due to shifts in technology), the value of these contracts may decrease or even become invalid.
- Continued Risk: Although the focus has shifted from Bitcoin prices to AI demand, there is still an element of uncertainty. If the AI industry declines, mining companies could end up with worthless assets.
Conclusion
In essence, the transition of mining companies is about pursuing new sources of profit. The logic remains the same: they follow where the money is. This time, the stakes are higher, as the focus has shifted to the long-term growth of the AI industry. Whether this transition will be successful depends on the sustainability of AI demand and the ability of mining companies to sustain their operations until they start earning rent from their new businesses. For now, they have found a more stable source of revenue than mining Bitcoin.