Burdened by 1.3 billion yuan in debt: Wu Jihan's Bitdeer is fighting an AI war with no way out.

Author: Rhythm

On February 20, 2026, BitDeer posted its weekly production update on X: 189.8 BTC were mined that week and sold. The remaining inventory of 943.1 BTC was sold in one go.

Bitcoin balance: 0.

In fact, Bitcoin mining has been a form of time arbitrage since day one.

Use today's electricity and machinery to trade for tomorrow's Bitcoin. The garage doesn't need processing, customers, or a brand. The investment is current costs; the bet is on future prices. If your judgment is correct, time will make you money.

This logic has been running for over a decade. What Wu Jihan is doing now is changing the target of this logic.

The target has shifted from the price of cryptocurrency to the long-term price of computing power demand under the AI ​​climate. The means have changed from exchanging electricity for cryptocurrency to borrowing money to buy land. The object of arbitrage has changed, but the structure of arbitrage has disappeared.

In the same week that it cleared out its Bitcoin holdings, BitDeer also priced out $325 million in new debt.

According to Bitdeer's financial report, as of December 31, 2025, Bitdeer had $1 billion in borrowings. Therefore, its total debt is approximately $1.3 billion.

The debt is real, the land purchase is real, but the outcome of this tough battle may not be revealed until 2029.

1. Mining companies that don't want to do AI are not good companies.

Founded in 2018, BitDeer started as a mining machine sharing platform. It is currently one of the world's largest listed mining companies, with a self-mining hashrate of 63.2 EH/s, ranking first among listed mining companies globally and accounting for approximately 6% of the entire Bitcoin network's hashrate.

But now, Wu Jihan doesn't want to sell computing power anymore; he wants to get into the power industry.

Looking at Bitdeer's financial statements, as of early 2026, Bitdeer's total global power pipeline capacity was 3002 MW, of which 1658 MW were already online and operational, and 1344 MW were under construction or planned. Microsoft and Google's individual hyperscale data centers typically range from 100 to 300 MW in size.

In other words, 3002 MW is equivalent to packaging the power requirements of 10 to 30 Google hyperscale data centers into one company. Therefore, Bitdeer's pipeline capacity is very impressive on paper.

The primary purpose of the $1.3 billion debt is to secure power land assets globally to pave the way for the transformation into an AI data center.

The first one is Rockdale, Texas, with a capacity of 563 MW (including a 179 MW expansion), which is already operational and primarily used for mining. This is their core business, and their cash flow is stable.

Secondly, the Clarington, Ohio project, a 570 MW facility with a 30-year lease and a signed power contract, was originally scheduled for completion in Q2 2027 and is positioned as an HPC/AI core site. This is the core of the entire AI transformation plan, and also the biggest potential pitfall, which we will discuss in detail later.

Next is Tydal in Norway, with a 175 MW mine, which is being converted into an AI data center, expected to be completed by the end of 2026, providing 164 MW of effective IT load. It has competitive hydropower resources and energy costs. The conversion cost is far lower than building a new facility. Currently, it's the fastest-progressing and least risky option.

Land, electricity, and data centers are known in the AI ​​industry as the "most difficult assets to replicate." BitDeer accumulated these through ten years of mining farm operation.

One rarely mentioned aspect deserves special mention: SEALMINER. BitDeer isn't just building data centers; it's also developing its own mining chips. The SEAL series has reached its third generation, with the SEAL03 achieving 9.7 joules per terahash. The A3 Pro, mass-produced in September 2025, has already entered the top tier globally. The SEAL04 aims for 5 joules per terahash; if achieved, it will surpass all mass-produced mining machines on the market. The gross profit margin of its self-developed chips exceeds 40%, far higher than mining itself.

This is him repeating what he did back in Bitmain: from buying other people's shovels to making his own.

II. How much was borrowed, and how much benefit can AI bring?

To pursue its AI ambitions, Bitdeer had borrowed over $1 billion by the end of 2025. Adding the $325 million in new debt in February 2026, its total debt exceeded $1.3 billion.

In less than two years, multiple rounds of financing were completed. In May 2024, Tether invested $100 million, becoming the second-largest shareholder, with warrants allowing for an additional $50 million. Three months later, the first convertible bond tranche of $150 million was issued, with an annual interest rate of 8.5%. In November of the same year, the second tranche of $360 million was issued, with the interest rate reduced to 5.25%.

In November 2025, a package deal was launched: 400 million convertible bonds plus 148.4 million shares issued, two supporting transactions. In February 2026, another 325 million convertible bonds plus 43.5 million shares were issued, while 135 million of the proceeds were used to repurchase the earliest batch of old bonds issued in 2029, pushing the repayment period to 2032.

The total exceeds $1.4 billion. The money flowed into mining rigs, data centers, and AI infrastructure, along with rolling debt rollovers.

However, every time Bitdeer issues bonds, its stock price drops by 10% to 17%. This has become a fixed market reaction. Fortunately, the company still manages to secure funding each time.

The core of the loan structure is convertible bonds. These new bonds, due in 2032, have an initial conversion price of approximately $9.93, a 25% premium over the concurrent equity offering price of $7.94. Once the stock price reaches that level, the bondholders will convert to shares, without needing cash. The company doesn't actually need to repay any money; it only needs the stock price to rise.

The logic behind convertible bonds is to seek a rise in their own stock price. This is inherently a gamble on whether the AI ​​narrative can be accepted by the market. The annual interest burden, calculated at an average interest rate of 5% and a principal of 1.3 billion, exceeds $65 million annually. Meanwhile, the total AI/HPC Cloud revenue for the entire year of 2025 is less than a fraction of six months' interest payments.

Currently, this interest is being rolled over entirely by issuing more bonds. It's impossible to say that the pressure isn't significant.

Given such a large investment, there must be a desire to see more substantial returns. So let's look at BitDeer; how much benefit can AI bring?

The AI ​​business currently earns $10 million a year, accounting for less than 2% of total revenue. For a company with a market value of nearly $2 billion, this figure is almost negligible.

Of course, this is not the end.

Bitdeer's GPUs tripled in three months, from 584 to 1792. Utilization dropped from 87% to 41%, mainly because the machines were being deployed too quickly; the B200/GB200 series was still in customer testing and hadn't started generating revenue yet. Power was on, machines were being installed, and the denominator (revenue) was skyrocketing, but revenue hadn't caught up.

How high is the ceiling?

Roth/MKM estimates that with full deployment of HPC capacity, the annualized revenue potential is $850 million. Management is even more aggressive: all 200 MW will be invested in AI cloud, generating over $2 billion annually, three times the total mining revenue for 2025.

However, both of these figures are contingent on three conditions: the construction is completed on schedule, a long-term contract at the hyperscaler level is secured, and the GPUs are running at full capacity.

None of the three conditions have been met yet.

This is the battle that BitDeer is currently fighting: mining to support AI, AI is making promises, and whether those promises can be realized depends on the execution over the next two or three years.

Third, the key challenge lies in how narrow the time window is.

$1.3 billion in debt sounds dangerous. But BitDeer's debt structure is designed to be more stable than it appears.

Highly leveraged companies often die for the same reason: a concentration of debt maturing, insufficient cash, and forced liquidation at a loss.

Bitdeer set the maturity dates for its three batches of convertible bonds in 2029, 2031, and 2032, respectively.

To some extent, this can be seen as a deliberately created buffer zone. By the time the first batch expires, Tydal and Clarington will theoretically be operational; by the time the second batch expires, AI revenue should already be speaking for itself; by the time the third batch expires, the market will have its own judgment on what this company truly is. Three milestones, three opportunities for renegotiation.

Convertible bonds provided time, but Wall Street wasn't buying it. Keefe Bruyette cut its price target from $26.50 to $14. The current share price is around $8. The market's signal is realistic: a transformation story needs to generate revenue.

But all this pressure gave Wu Jihan what he needed most, and also what was most cruel: time.

The smooth path might unfold like this: By the end of 2026, Tydal's transformation is complete, the 164 MW hydroelectric data center in Norway is online, and contracts with European clients begin to come in. In 2027, Clarington wins the lawsuit, the 570 MW Ohio project officially breaks ground, and major US clients follow suit. By 2028-2029, both core assets are operating at full capacity, revenue is approaching $1 billion, and analysts are reclassifying Bitdeer from a discounted mining company to a premium AI infrastructure provider. In 2029, the first batch of bonds matures, and bondholders, watching the stock price, will most likely choose to exchange for shares rather than cash.

In every tough battle, Wu Jihan had to time it perfectly.

Then came Clarington.

In the same industrial park in Ohio, a steel manufacturer called American Heavy Plate Solutions signed a 30-year lease on 9.9 acres of land in 2018. They sued BitDeer, arguing that building an AI data center would interfere with shared power, roads, railroads, and communication lines, violating restrictive clauses. Their request is for a permanent injunction to prevent BitDeer from commencing construction.

Clarington represents 42% of the pipeline under construction. If it gets stuck, the entire timeline will have to be rewritten.

Therefore, Bitdeer's biggest single risk is not its debt or its stock price, but its steel plant.

The mining industry hasn't been idle either. In February 2026, the Bitcoin network difficulty surged by 14.7%, the largest single jump since May 2021. With the same electricity costs, fewer coins were mined. The Q4 gross margin had already dropped from 7.4% a year ago to 4.7%. The mining business is slowly thinning.

The worst-case scenario is also clear: the Clarington lawsuit drags on for two years, and construction is halted; Tydal is delayed, and GPU utilization continues to hover from 41%; the first batch of bonds will mature in 2029, and there is not enough cash on hand, forcing refinancing, further diluting the stock price, and making it increasingly difficult to reach the conversion threshold.

Both paths truly exist.

Fourth, sell all your Bitcoin, then what?

There's a tradition in the mining community: hoarding coins is a matter of faith, an endorsement of Bitcoin's long-term value.

MARA has accumulated 53,250 BTC, Riot has accumulated 18,000, and Strategy has accumulated 710,000. The more you accumulate, the more the market believes in you.

BitDeer is currently at zero.

The official explanation is: selling cryptocurrency is to provide liquidity for land purchases. That makes sense. Other companies in the industry are moving in the same direction: Riot sold $200 million worth of Bitcoin for AI expansion, Bitfarms is abandoning its "Bitcoin company" image, and MARA is also investing in HPC.

But there's something more fundamental than identity iteration.

From day one, the mining industry has been betting on the same thing: that something in the future will be more expensive than it is today. Ten years ago, mining was betting on a rise in cryptocurrency prices. Now, buying land is betting on an explosion in demand for computing power.

The target may change, but the logic of time arbitrage remains the same.

What Wu Jihan truly bought was a position where "no matter who wins, they have to pay my electricity bill."

They don't bet on the entire market; they just control the entrance. Amazon didn't bet on which internet company would win; they simply rented out their servers to everyone. AT&T doesn't care what you talk about on the phone; they only care if you make a call.

From selling products to selling services to collecting rent, the evolution of industries has always followed only one path.

The only difference is whether you walk there voluntarily or are pushed there by someone.

Wu Jihan bought this window of opportunity for over a billion US dollars. He's waiting for the AI ​​funding to catch up with the speed of his debt.