I have recently noticed an interesting movement in the mining world. Bitdeer is starting to sell all of its Bitcoin reserves completely. This is not just unusual; it is doing so to fund a massive project: transitioning from cryptocurrency mining to AI infrastructure.

Last February, Bitdeer sold 189.8 Bitcoin from that week’s production and fully depleted its reserves. At the same time, it closed a new financing round worth $325 million. When all numbers are added up, the total debt now exceeds $1.3 billion. That’s not a small amount.

The logic behind this move is clear: founder Wu Jihan is betting on something different this time. Previously, mining bets on rising Bitcoin prices. Now, the bet is on the increasing demand for computing power in the age of artificial intelligence. The goal has changed, but the core idea remains: betting on the future.

The assets Bitdeer accumulated from years of mining are not simple. It now has 3,002 megawatts of global energy capacity, of which 1,658 megawatts are already operational and 1,344 megawatts are under construction. This is equivalent to the energy needs of 10 to 30 giant data centers like Google or Microsoft in just one company.

There are three main projects. First, Rockdale in Texas with 563 megawatts, already operational and focused on traditional mining. Second, Clarington in Ohio with 570 megawatts, which is the real heart of the project. Third, Tydal in Norway, converting an old mining facility into an AI data center with 164 megawatts of cheap hydropower.

On the technical side, Bitdeer has developed its own chips called SEALMINER. The third generation, SEAL03, has reached an efficiency of 9.7 joules per terahash. The fourth generation, SEAL04, aims for 5 joules per terahash, and if achieved, it will outperform all current mining equipment on the market. The profit margin on these chips exceeds 40%, much higher than mining profits themselves. This is similar to what Wu Jihan did previously at Bitmain: from buying others’ tools to manufacturing his own.

But current AI numbers are very modest. Revenue from AI operations is currently around $10 million annually, less than 2% of total revenue. The number of GPU units increased from 584 to 1,792 in just three months, but utilization rates dropped from 87% to 41%. The simple reason: equipment is being installed quickly, but customer contracts have not yet started.

According to analyst estimates, if the capacity is fully utilized, annual revenue could reach $850 million. Management expects much more: $2 billion annually if 200 megawatts are fully dedicated to AI. But these figures depend on three conditions: completing construction on time, securing long-term contracts with major hyperscalers, and fully utilizing GPU units. So far, none of these have been achieved.

The annual interest burden, based on an average interest rate of 5% on $1.3 billion, exceeds $65 million per year. That’s real pressure. The company relies on issuing new bonds continuously to keep things moving. Every time bonds are issued, the stock drops 10% to 17%. This has become an automatic market reaction.

The biggest risk now is not the debt itself but the details of execution. A steel factory called American Heavy Plate Solutions has filed a lawsuit against Bitdeer regarding the Clarington project, claiming that construction conflicts with shared infrastructure. Clarington accounts for 42% of the pipelines under construction. If it stops, the entire schedule collapses.

In the mining sector itself, pressure is increasing. In February, Bitcoin network difficulty rose 14.7%, the largest increase since May 2021. Profit margins fell from 7.4% to 4.7% in the last quarter. The mining path is becoming increasingly narrow.

The worst-case scenario is clear: a two-year delay in the Clarington lawsuit, Tydal’s postponement, GPU utilization remaining at 41%, the first bond maturity in 2029 with liquidity shortages, forcing the company to seek additional funding and continuous stock dilution.

But there is a deeper logic here. Wu Jihan didn’t buy products or services. He bought infrastructure. He bought land, electricity, and servers. He doesn’t care who uses them; what matters is that someone pays the electricity bill. This is not an investment in the winner but a monopoly on the road to the winner.

Amazon didn’t invest in a specific internet company; it rented servers for everyone. AT&T doesn’t care about the content of calls, only that the call is made. The long-term industry trend is only one: from selling products to selling services to collecting rent.

The question now is: can Wu Jihan reach the finish line before liquidity runs out? He has a very narrow window. He must complete Tydal by the end of 2026, win Clarington in 2027, and operate both projects at full capacity by 2029. If successful, revenues will jump to a billion dollars, and the company’s valuation will shift from a discounted mining firm to an AI infrastructure company with a premium. If he fails, pressure will become unbearable. Time is the most valuable currency now.