Just caught something worth thinking about. BitDeer cleared out all its Bitcoin holdings last week—sold 943.1 BTC in one go, leaving zero balance. At the same time, Wu Jihan's company priced out another $325 million in new debt. Total debt load now sits around $1.3 billion.

This isn't just a pivot; it's a bet. And honestly, it's a fascinating one.

For over a decade, Bitcoin mining was pure time arbitrage—use today's electricity to trade for tomorrow's Bitcoin. Simple math. But Wu Jihan is rewriting the playbook. He's shifting the bet from crypto prices to something bigger: the long-term demand for computing power in the AI era. The means have changed too—instead of converting electricity into coins, he's borrowing heavily to acquire land and power infrastructure globally.

Look at the numbers. BitDeer's global power pipeline capacity sits at 3,002 MW. To put that in perspective, that's equivalent to 10 to 30 hyperscale data centers like Google's. On paper, it looks impressive. The core plays are Rockdale, Texas (563 MW, already operational), Clarington, Ohio (570 MW, the crown jewel of the AI transformation), and Tydal, Norway (175 MW, converting from mining to AI data center).

Here's where it gets interesting. Wu Jihan isn't just building infrastructure; he's also developing proprietary mining chips through SEALMINER. The SEAL03 already ranks among the best globally, and the upcoming SEAL04 aims for 5 joules per terahash—potentially the most efficient on the market. Gross margins on these chips exceed 40%, way higher than mining itself. He's basically repeating what he did at Bitmain: from buying shovels to making them.

But the debt structure is where things get precarious. With an average interest rate around 5% on $1.3 billion, that's over $65 million in annual interest payments. Current AI cloud revenue? Less than $10 million annually. The math doesn't work yet. He's essentially betting that GPU deployment and revenue will catch up before the convertible bonds mature in 2029, 2031, and 2032.

The real risk isn't the debt itself—it's execution. Clarington represents 42% of the pipeline under construction. A lawsuit from a neighboring steel manufacturer could halt the entire Ohio project. If that stalls, the whole timeline collapses. Tydal's on track and lowest risk. But Clarington? That's the wildcard.

Meanwhile, mining margins are compressing. Bitcoin network difficulty jumped 14.7% in February alone—the largest spike since May 2021. With the same electricity costs, fewer coins. Q4 gross margins dropped from 7.4% to 4.7% year-over-year.

What Wu Jihan truly bought is leverage on computing demand. Amazon didn't bet on which internet company would win; they just rented servers to everyone. AT&T doesn't care what you say on the phone; they care if you make a call. He's positioning BitDeer to be the power and infrastructure provider in the AI race. No matter who wins, someone has to pay the electricity bill.

The question is whether the AI funding cycle moves fast enough to cover the debt maturity dates. By end of 2026, if Tydal's operational. By 2027, if Clarington clears legal hurdles. By 2028-2029, if both core assets are running full capacity with enterprise contracts. Three milestones, three chances to refinance or convert bonds to equity.

It's a high-wire act. The window is narrow, and the margin for error is thin. But if he pulls it off, the narrative flips from "distressed mining company" to "AI infrastructure powerhouse." That's worth watching closely over the next couple of years.