Bitdeer's $1.3 Billion Debt and the Mining Transformation Dilemma of Riot and Others

At a pivotal point in cryptocurrency mining history, major mining companies like Bitdeer and Riot are making high-stakes strategic shifts. In February 2026, Bitdeer completed a new round of financing, with total debt surpassing $1.3 billion—far exceeding the liabilities of competitors like Riot and MARA. This is not just a financial crisis but a microcosm of the industry’s shift from Bitcoin mining to AI infrastructure.

Once, mining was simple: use today’s electricity and mining rigs to earn tomorrow’s Bitcoin. Now, companies like Bitdeer are rewriting that logic—buying up global power assets with massive debt, betting that future AI computing demand will pay off. While Riot, MARA, and others are also dabbling in AI, their strategies are more conservative, maintaining Bitcoin holdings. Which path is better remains to be seen.

From Mining Hardware to Power Empire: Bitdeer’s Transformation

Founded in 2018 as a mining hardware sharing platform, Bitdeer has become one of the world’s largest publicly listed mining companies. Latest data shows its mining hash rate at 63.2 EH/s, about 6% of the global Bitcoin network—ranking first among listed miners. But that “number one” badge is fading.

Bitdeer’s true goal has shifted. CEO Wu Jihan no longer aims to sell hash rate but to control electricity. By early 2026, Bitdeer’s global power capacity reached 3,002 MW, with 1,658 MW in operation and 1,344 MW under construction or planned. This is equivalent to the power needs of 10 to 30 Google data centers.

Behind this ambition lies enormous debt: by the end of 2025, Bitdeer’s liabilities exceeded $1 billion, and with an additional $325 million in convertible bonds issued in February 2026, total debt surpassed $1.3 billion. Most of this capital is directed toward acquiring power assets worldwide, paving the way for AI data center transformation.

Bitdeer’s power projects include three core assets: a 563 MW facility in Rockdale, Texas (including a 179 MW expansion), already operational; a 570 MW project in Clarington, Ohio, with a 30-year lease, positioned as an HPC/AI hub, expected to be completed by Q2 2027—central to the transformation but also the riskiest; and a 175 MW hydroelectric project in Tydal, Norway, converting from mining to AI data center, expected to finish by late 2026, offering the best cost and lowest risk.

Beyond power assets, Bitdeer is also making strides in chip development. Its self-developed SEALMINER series has reached the third generation, with SEAL03 achieving 9.7 J/T, and the A3 Pro in testing, now among the top miners globally. Future SEAL04 aims for 5 J/T, surpassing all mainstream miners. Chip margins exceed 40%, far higher than mining itself. This echoes Wu Jihan’s earlier strategy at Bitmain—evolving from buying tools to manufacturing them.

The Gap Between AI Promises and Reality

To push this transformation, Bitdeer has completed multiple funding rounds in less than two years. In May 2024, Tether invested $100 million, becoming the second-largest shareholder, with warrants for an additional $50 million. Followed by two convertible bond rounds: $150 million at 8.5% annual interest, and $360 million at 5.25%. By November 2025, Bitdeer raised even more—issuing $400 million in convertible bonds and 14.84 million new shares. In February 2026, it raised another $325 million in convertible bonds and 4.35 million shares, repurchased $135 million of old bonds, extending maturities to 2032. Total financing exceeded $1.4 billion.

What has this influx of capital bought? By the end of 2025, Bitdeer’s AI/HPC cloud business generated only $10 million annually—less than 2% of total revenue. According to financial reports, annual interest expenses already exceeded $650 million (assuming 5% average interest on $1.3 billion debt). This means current interest costs are sustained only by issuing new debt.

GPU deployment has skyrocketed, but efficiency has declined. Over three months, GPU count grew from 584 to 1,792—tripling—yet utilization dropped from 87% to 41%. The main reason: rapid deployment outpaced testing, and new GPUs like B200/GB200 are still in testing phases, not yet generating revenue. This is a classic “burn money first, profit later” model.

Analysts Roth/MKM conservatively estimate that if HPC capacity is fully utilized, annual revenue could reach $850 million. More aggressively, Bitdeer’s management assumes dedicating 200 MW entirely to cloud AI could generate over $2 billion annually—three times its 2025 mining revenue. But both scenarios depend on three conditions: on-time construction, securing long-term contracts with major clients, and full GPU utilization—none of which are currently met.

In contrast, Riot, MARA, and others still hold significant Bitcoin reserves—Riot owns 18,000 BTC, MARA 53,250 BTC—while Bitdeer has liquidated all its Bitcoin holdings. The official reason: to improve liquidity for power asset acquisitions. But this signals a belief that future value of power assets will outstrip Bitcoin itself.

Time Window and Overhanging Risks

Bitdeer’s debt structure appears carefully designed. The three convertible bonds mature in 2029, 2031, and 2032—seemingly planned with a buffer. The idea: when the first matures, Tydal and Clarington will be operational; when the second matures, AI revenue should be evident; and by the third, the market will have given its final verdict. Each maturity is an opportunity for renegotiation.

However, Wall Street is skeptical. Keefe Bruyette lowered its target price from $26.50 to $14, and the current stock hovers around $8. The market signals: the transformation story must deliver real revenue.

Wu Jihan faces dual pressures—what he needs most is time, but time is also the cruelest factor. An optimistic scenario might look like this:

By late 2026, Tydal’s renovation completes, a 164 MW hydroelectric plant in operation, European clients start placing orders. In 2027, Clarington wins a legal case, Ohio’s 570 MW project begins construction, and major US clients follow. Between 2028 and 2029, these core assets operate at full capacity, with revenues soaring into the billions. Analysts shift their view, moving from “mining discount” to “AI infrastructure premium.” When the first debt maturities arrive in 2029, creditors, seeing the stock price and prospects, are likely to convert debt into equity.

But a more immediate threat looms. In Ohio’s industrial park, a steel producer called American Heavy Plate Solutions signed a 30-year land lease in 2018. They sued Bitdeer, claiming that AI data center construction interferes with shared power, roads, railways, and communications, violating restrictions. They seek a court order to permanently block Bitdeer’s construction.

Clarington accounts for 42% of Bitdeer’s planned pipeline capacity. If litigation drags on for two or three years, the entire timeline must be rewritten. Currently, Bitdeer’s largest risk isn’t debt or stock price fluctuations but a lawsuit from an Ohio steel plant.

Mining operations are also not immune. In February 2026, Bitcoin network difficulty surged 14.7%, the largest single jump since May 2021. Under the same electricity, fewer coins are mined. Q4 2025 gross margins fell from 7.4% a year earlier to 4.7%. Traditional mining profits are thinning.

The worst-case scenario is clear: two or three years of litigation delay in Ohio, Tydal’s project postponed, GPU utilization stuck at 41%, and by 2029, when bonds mature, cash reserves may be insufficient for repayment, forcing refinancing and further dilution. Both paths are real possibilities.

Power Play: A New Chapter in Time Arbitrage

In mining history, a tradition exists: holding coins reflects faith—supporting Bitcoin’s long-term value. Marathon holds 53,250 BTC, Riot 18,000 BTC, Strategy even 710,000 BTC. The more they hold, the more the market believes they are bullish. Bitdeer, however, has sold all its Bitcoin.

Officially, the reason is to gain liquidity for power asset acquisitions. That’s true. Peers are heading the same way—Riot uses $200 million of Bitcoin proceeds to expand AI infrastructure; Bitfarms is shedding its “Bitcoin company” label; MARA is building its HPC ecosystem.

But deeper logic warrants reflection. From the start, the mining industry bet on one thing: that the future price of something will be higher than current costs. Ten years ago, it was Bitcoin price appreciation. Now, it’s buying power assets betting on AI compute demand explosion. The object has changed, but the core logic of time arbitrage remains.

Wu Jihan’s real play is positioning himself as: “Whoever wins, they pay me electricity.” No need to guess which path—just control the intersection. Amazon doesn’t bet which internet company will win; it rents servers to everyone. AT&T doesn’t care what you say on the phone, only whether you make calls. From selling products to services to rent collection, industry evolution follows a single direction. The only difference: whether you choose to walk that path or are pushed onto it.

Wu Jihan has spent billions to buy this time window. Now, he’s waiting for AI revenue to catch up with debt growth. Compared to conservative players like Riot and MARA, Bitdeer’s path is more aggressive—and riskier. The answer will come in 2029.