Wu Jihan's BitDeer: Betting $1.3 Billion on AI's Power Demand

Wu Jihan is making a calculated wager that most mining executives would consider reckless. On February 20, 2026, BitDeer published its latest production metrics: 189.8 BTC mined and liquidated within a single week. The remaining inventory of 943.1 BTC followed immediately after. Current Bitcoin holdings: zero. This wasn’t forced liquidation—it was deliberate capital reallocation. The message was unmistakable: BitDeer is no longer in the business of accumulating digital assets. It’s reorganizing itself entirely around infrastructure and artificial intelligence.

The pivot makes sense historically. Mining has always functioned as a form of temporal speculation. You deploy today’s capital—electricity and hardware—to capture tomorrow’s asset appreciation. For over a decade, that equation worked predictably: cheaper power multiplied into higher-priced Bitcoin. Wu Jihan is applying that same arbitrage logic to a different market. Instead of betting on cryptocurrency valuations, he’s positioning BitDeer as a critical node in AI infrastructure, where computing power demand may dwarf even cryptocurrency’s trajectory.

But the arithmetic is stark. By the end of 2025, BitDeer had accumulated over $1 billion in debt obligations. February 2026 added another $325 million in convertible bond issuance. Total outstanding debt now exceeds $1.3 billion. The company has mortgaged itself to acquire what the AI industry calls “unreplicable assets”: energy resources, land rights, and data center capacity distributed across multiple continents.

From Mining Dominance to Infrastructure Scale

BitDeer entered the market in 2018 as a mining hardware-sharing platform. Its trajectory has been remarkable. By early 2026, the company operated the largest self-mining hashrate among publicly listed mining firms: 63.2 exahashes per second, representing approximately 6% of Bitcoin’s entire network compute power. That positioning alone would justify a substantial valuation. But Wu Jihan has decided it’s insufficient.

The reason lies in BitDeer’s evolving power footprint. As of early 2026, the company controlled or held contracts for 3,002 megawatts of global power capacity—roughly equivalent to combining the infrastructure requirements of 10 to 30 of Google’s hyperscale data centers into a single entity. Of this total, 1,658 MW were operational; 1,344 MW remained under construction or in development. This isn’t mining company scale anymore. It’s infrastructure company scale.

The three flagship projects reveal the strategy’s architecture. Rockdale, Texas presents a proven core: 563 MW of capacity (including a 179 MW expansion), already generating stable cash flow from cryptocurrency mining. Tydal, Norway offers an elegant conversion opportunity: a 175 MW mining facility being transformed into an AI data center with competitive hydroelectric costs. The transformation is expected by year-end 2026, delivering 164 MW of effective compute load.

Then there’s Clarington, Ohio—the crown jewel and the potential catastrophe. A 570 MW facility with 30-year power contracts, theoretically scheduled for completion in Q2 2027, positioned as the centerpiece of BitDeer’s North American AI infrastructure ambitions. This single project represents 42% of BitDeer’s construction pipeline. If delayed, the entire financial structure collapses.

The Debt Architecture: Convertible Bonds and Market Bets

Understanding BitDeer’s finances requires grasping one instrument: convertible bonds. Between May 2024 and February 2026, Wu Jihan executed a sophisticated fundraising campaign through multiple convertible issuances. The earliest tranches carried 8.5% interest rates; later iterations reduced to 5.25%. The latest batch, issued in February 2026, matures in 2032 at a conversion price around $9.93—approximately 25% above concurrent equity pricing at $7.94.

This structure reveals the implicit wager: BitDeer isn’t primarily betting on revenue generation. It’s betting on stock price appreciation. As long as the share price rises above the conversion threshold, bondholders voluntarily convert their debt into equity, eliminating repayment obligations. The company doesn’t need to generate sufficient cash to redeem $1.3 billion—it just needs the equity markets to believe in the AI infrastructure thesis.

The annual interest burden, calculated conservatively at 5% against $1.3 billion in principal, exceeds $65 million yearly. Meanwhile, BitDeer’s entire AI and HPC cloud segment generated less than $10 million in 2025 revenue—a negligible fraction of debt servicing costs. Currently, interest payments are themselves being financed through successive bond issuances, creating a rolling refinancing cycle that markets have grown accustomed to accepting.

Predictably, Wall Street has expressed skepticism. Keefe Bruyette downgraded its price target from $26.50 to $14 following the repeated equity dilution from successive financing rounds. The current share price hovers around $8, well below conversion requirements. Yet the market continues to fund successive tranches, suggesting that even skeptical capital believes some version of BitDeer’s AI narrative will eventually validate the valuation.

The Hardware Innovation: SEAL Chips and Competitive Advantage

BitDeer’s competitive moat extends beyond site selection and power contracts. The company is vertically integrating chip design through its proprietary SEAL series. This recalls Wu Jihan’s Bitmain playbook: transition from purchasing others’ equipment to manufacturing your own.

The progression is instructive. SEAL series chips have reached their third generation. The SEAL03 achieved 9.7 joules per terahash efficiency—a competitive benchmark globally. The mass-produced A3 Pro, launched in September 2025, already ranks among the world’s top-tier mining hardware. The unreleased SEAL04 targets 5 joules per terahash. If achieved, it would surpass all mass-production competitors currently available.

The margin structure explains the strategic priority. Gross margins on self-developed chips exceed 40%—substantially higher than mining operations themselves. For a company transitioning away from direct mining, these margins matter disproportionately. They provide cash generation decoupled from Bitcoin price movements or network difficulty adjustments.

GPU Deployment: The Velocity Mismatch Problem

BitDeer’s GPU inventory reveals the operational intensity of infrastructure scaling. In just three months, GPU counts expanded from 584 units to 1,792—a tripling of capacity. Simultaneously, utilization rates collapsed from 87% to 41%. The machines arrived faster than revenue could follow.

The reason is straightforward: B200 and GB200 series units were still in customer evaluation phases, not yet generating commercial revenue. Power infrastructure was activated, equipment was installed, but customer contracts hadn’t materialized. This creates a temporary but severe cash pressure. Revenue numerators are static while denominators (installed capacity) multiply.

Analyst projections offer competing scenarios. Roth and MKM estimate that full HPC capacity deployment could generate approximately $850 million in annualized revenue. BitDeer management is more aggressive: allocating 200 MW fully to AI cloud could exceed $2 billion in annual revenue—roughly three times all cryptocurrency mining revenue for 2025. These projections rest on three prerequisites: on-time construction completion, enterprise-level long-term contracts, and full GPU utilization. None have been achieved yet.

The Clarington Obstacle: When Infrastructure Meets Litigation

Wu Jihan’s grand narrative encountered an unexpected antagonist in early 2026: American Heavy Plate Solutions, a steel manufacturer located in the same Ohio industrial park as Clarington. The company holds a 30-year lease on 9.9 acres, signed in 2018, long before BitDeer’s arrival.

The lawsuit claims that a 570 MW AI data center would disrupt shared power infrastructure, roads, railroads, and communication facilities—violating restrictive covenants embedded in the original lease agreement. American Heavy Plate Solutions seeks a permanent injunction preventing construction commencement.

The stakes are enormous. Clarington represents 42% of BitDeer’s entire construction pipeline. A prolonged legal battle—say, two years—would obliterate the financial timetable. Bond maturity schedules assume Clarington operationalizes by 2027. If construction remains blocked, no amount of Tydal success or GPU efficiency improvements can offset the lost capacity.

This is why industry observers have concluded that BitDeer’s primary risk isn’t debt volatility or stock price compression—it’s a steel plant in Ohio.

The Narrowing Time Window: 2029 as the Deadline

The debt structure deliberately spaces maturity dates across 2029, 2031, and 2032. This creates what management characterizes as deliberate buffer zones. By 2029, Tydal should be operational and generating European client contracts. By 2031, Clarington presumably functions at scale, supplying major North American hyperscalers. By 2032, the market will have rendered its judgment on whether BitDeer truly became an AI infrastructure powerhouse or remained essentially a declining mining company.

But these timelines face mounting pressure. February 2026 brought a 14.7% surge in Bitcoin network difficulty—the largest single-month increase since May 2021. Mining profitability compressed accordingly. BitDeer’s Q4 gross margins had already declined from 7.4% to 4.7%. The legacy mining business isn’t generating enough cash to absorb infrastructure construction costs or debt servicing.

The negative scenario is equally vivid: Clarington litigation extends across 24+ months; Tydal experiences deployment delays; GPU utilization remains depressed; 2029 bond maturity arrives with insufficient cash on hand, forcing refinancing at worse terms, further diluting equity value, making conversion thresholds increasingly difficult to achieve.

The Philosophical Gamble: Who Pays the Electric Bill?

Perhaps the most revealing moment in BitDeer’s strategic evolution came when Wu Jihan zeroed out Bitcoin holdings entirely. In cryptocurrency mining culture, accumulation represents faith—a signal to the market that leadership believes in digital asset appreciation. Marathon Digital accumulated 53,250 BTC. Riot accumulated 18,000. Bit-Strategy accumulated 710,000. BitDeer is now at zero.

The official rationale—selling to fund land purchases—is technically accurate but incomplete. More fundamentally, the move reveals a philosophical reorientation. For the past decade, mining executives bet that something future would appreciate relative to present value. Bitcoin mining represented a wager on cryptocurrency price trajectories. Infrastructure acquisition represents a wager on computational demand explosions.

The underlying arbitrage logic, however, remains identical. Wu Jihan isn’t trying to predict cryptocurrency winners or AI model winners specifically. He’s constructed a position where “regardless of which technology dominates, someone must pay my electricity bill.” This echoes Amazon’s infrastructure play—not betting which internet companies would succeed, merely renting servers to all of them. It mirrors AT&T’s indifference about phone call content—only charging whether calls occurred.

The corporate evolution path has historically followed only one direction: from selling products to offering services to collecting rent. Whether companies walk this path voluntarily or are pushed into it by competition remains the only variable. Wu Jihan paid over $1.3 billion for a position along that path. He’s now waiting to discover whether AI’s funding velocity can match his debt obligations’ acceleration.

The years between 2026 and 2029 will determine whether this wager validates that investment or exposes it as catastrophically mistaken.