So, Bitdeer has run out of Bitcoin stock. It sounds strange coming from a mining company, but there’s logic behind it. In early February, they produced 189.8 BTC, sold everything, and the balance was zero. This generated quite a buzz in analyst communities — some people are discussing in Telegram groups what it means to make money from this transition.

But the point that few are seeing is deeper: Wu Jihan is not abandoning mining, he’s pivoting his bet. For ten years, the game was simple — you buy electricity and machines today to sell Bitcoin tomorrow. Now it’s changed. The currency he wants isn’t Bitcoin, it’s processing capacity under the wave of AI. And for that, he needs land, energy, and data centers.

The numbers speak: by the end of 2025, Bitdeer had 1 billion in accounting loans. In February 2026, they closed another 325 million in convertible bonds. Total: 1.3 billion in debt. All of this to lock down global assets. Rockdale in Texas (563 MW), Clarington in Ontario (570 MW), Tydal in Norway (175 MW) — it’s like combining the energy demand of 10 to 30 giant Google data centers into one company.

The AI business still generates only 10 million per year. Insignificant. But GPUs are being installed rapidly — they went from 584 to 1,792 in three months. Utilization dropped to 41% because the machines are still in testing phase. It’s the typical growth situation: the denominator is rising fast, revenue still isn’t keeping up.

Analysts estimate that when everything is operational, AI revenue could reach between 850 million and 2 billion per year. But that depends on three things: building on time, long-term contracts with giant hyperscalers, and GPUs running at full capacity. None of these three conditions have been met yet.

And there’s one detail few mention: Bitdeer developed its own mining chips, the SEAL series. The SEAL03 has an efficiency of 9.7 joules per terahash, and they’re aiming for the SEAL04 with 5 joules per terahash. The gross margin on the chips exceeds 40% — well above pure mining. It’s like what Wu Jihan did at Bitmain: moving from buying third-party tools to manufacturing their own.

Now comes the complicated part. The debt structure was designed with maturity dates in 2029, 2031, and 2032. It’s a conscious strategy to create a buffer. But the market isn’t paying for that — the target price was cut from $26.50 to $14, and the stock is around $8. The message is clear: the story of transformation needs to generate real revenue.

The ideal scenario would be: by the end of 2026, Tydal comes online with 164 MW hydroelectric generating revenue in Europe. In 2027, Clarington wins the legal process (yes, a steel mill is suing Bitdeer for interference in shared infrastructure), and the 570 MW project is officially launched. Between 2028 and 2029, both assets operate at full capacity, revenue rises to around 1 billion, and analysts reclassify the company from a discounted mining firm to an AI infrastructure company with a premium.

But there’s a worse scenario too: Clarington drags on for two years, Tydal faces delays, GPU utilization remains at 41%, bonds mature in 2029 without enough cash, forcing a new raise that further dilutes shares. Two roads, both real.

What Wu Jihan really bought with these billions was a position: no matter who wins in AI, someone will need to pay the electricity bill. It’s not a bet on which company will dominate, it’s a bet on controlling the entry point. Amazon didn’t bet on which internet startup would win, they just rented servers. AT&T doesn’t care what you say on the phone, only that you called. Evolution always follows the same path: from selling products to selling services, then to charging rent.

The question now is whether Wu Jihan can keep the timing. If the debt can be rolled over while the data centers start generating real revenue. The window is narrow, and any delay in Clarington or Tydal changes the entire timeline.