CleanSpark just sold $1.15 billion of zero-coupon convertible notes to buy more power and machines in the most brutal mining environment yet.
The deal is a 144A private placement due in 2032, with an initial conversion price of around $19.16, roughly a 27.5% premium to the $15.03 stock price at the time of announcement.
Approximately $460 million is allocated directly to repurchasing CleanSpark shares from the note buyers, with the remainder used to expand power and land portfolios, build data center infrastructure, including AI and high-performance computing capacity, repay Bitcoin-backed credit lines, and cover general corporate expenses.
That single transaction is a cheat sheet for where miner economics stand in 2025. The terms reveal who survives, who gets consolidated, and what it actually costs to stay relevant in a network that has just crossed one zettahash per second of global hashrate.
Whether the bet pays off depends less on narrative and more on whether cash flows can support a balance sheet that now carries over $1.7 billion in long-term debt against a treasury of roughly 12,100 BTC.
Zero percent means something
A zero-coupon convertible note of this size suggests that credit investors are comfortable being paid in equity optionality rather than cash interest.
They’re betting that CleanSpark will remain solvent despite multiple difficulties and price cycles, and maintain sufficient liquidity in equity for eventual conversion.
That’s a cost-of-capital advantage compared to smaller miners, which often resort to expensive equity dilution or high-yield debt with double-digit coupons. In 2025, only the most efficient miners can borrow this much at zero percent. Everyone else is paying up or getting consolidated.
However, the structure carries risk. It’s a leveraged bet on both Bitcoin price and CleanSpark equity performance. If execution stumbles or Bitcoin underperforms, the converts become a delayed dilution bomb.
If the stock trades well above $19.16, existing shareholders get diluted as note holders convert. The share buyback complicates the picture further, as CleanSpark is using $460 million of borrowed money to repurchase its own stock from the same investors buying the notes.
That signals management thinks the equity is undervalued, but it also means less capital available for actual expansion. After the buyback, approximately $670 million remains for capital expenditures and debt repayment.
Capex and scale in a one-zettahash world
New-generation mining rigs, along with their associated infrastructure, typically cost between $6 million and $10 million per exahash per second of capacity.
If CleanSpark deployed all incremental capital into mining, which is unlikely given the focus on AI and data centers, that $670 million could fund 70 to 110 exahashes of additional capacity.
In a network already above 1,000 exahashes, even half that would cement CleanSpark as a top-tier hashrate player.
A meaningful chunk also flows into power sites and AI or HPC build-outs, but the signal is clear: 2025 miner economics are now “go big or get eaten.”
Capital intensity is exploding beyond just buying rigs. Miners are building vertically integrated power and data center campuses, treating hashpower as part of a broader infrastructure play rather than a standalone bet on block rewards.
CleanSpark concluded its fiscal second quarter with approximately 42.4 exahashes per second and a stated goal of surpassing 50 exahashes by 2025, representing around 4.9% of the global hashrate at current levels.
The raise positions them to push further, but it also highlights the “treadmill” problem. The network hashrate continues to climb, the difficulty adjusts upward, and each exahash generates fewer Bitcoins over time.
Post-halving and post-one-zettahash, staying in place requires constant reinvestment to maintain revenue per unit of capacity.
Post-halving margin stack
CleanSpark’s fiscal numbers for the second quarter show revenue up 62.5% year-over-year to $181.7 million, but a net loss of $138.8 million and negative adjusted EBITDA. Cost to mine came in around $42,700 per Bitcoin, positioning them on the efficient end of the curve.
At roughly $103,000 Bitcoin, that implies a gross mining margin around 55% to 60% before selling, general and administrative expenses, interest, hosting, and other overhead.
Energy costs alone accounted for 46% of Bitcoin’s revenue in the second quarter.
That’s the post-halving reality: block subsidy halved, network hashrate at all-time highs, and hashprice compressed to levels that squeeze everyone but the most efficient operators.
Only miners with cheap, stable power, meaningful scale, and access to low or zero-coupon capital can keep positive margins after fixed costs.
The 2024 halving didn’t kill miners outright, but bifurcated them instead. CleanSpark’s raise says which side of that divide it intends to occupy.
Smaller miners without locked-in power deals or efficient fleets are either shuttering sites, selling assets, or raising dilutive equity through at-the-market programs.
CleanSpark is doing the opposite, raising debt-like capital with a simultaneous buyback, signaling confidence that future hashrate and Bitcoin holdings justify current equity valuations.
AI side quests: diversification or narrative sugar?
CleanSpark’s use of proceeds explicitly includes “data center infrastructure” and AI or HPC capacity. That language mirrors a broader industry trend as Core Scientific, Iris Energy, Hut 8, and TeraWulf pitch HPC and AI hosting as higher-margin uses for their power and infrastructure.
The market has grown skeptical of “AI pivot” slides without signed contracts and transparent unit economics.
The framework to judge whether this is real diversification comes down to the revenue structure. Will the AI builds be contracted, dollar-denominated, multi-year agreements that de-risk revenue? Or is this “we might host AI someday” optionality that competes with Bitcoin mining for capital but doesn’t deliver near-term cash flows?
AI and HPC hosting can generate steady, predictable revenue if appropriately contracted. However, those dollars compete directly with the incremental Bitcoin mined per megawatt, as well as the optionality value of holding self-mined Bitcoin in the treasury.
Every dollar CleanSpark spends building AI capacity is a dollar not deployed into hashpower, and the return profile is fundamentally different.
Bitcoin mining offers leveraged exposure to Bitcoin price appreciation. AI hosting offers utility-like revenue with lower volatility but also lower upside.
Separating narrative from cash flows
The pro forma capital stack now includes roughly $640 million in existing debt, plus $1.15 billion in new convertible debt, against equity, and a Bitcoin treasury worth approximately $1.25 billion at $103,000 per Bitcoin.
No interest expense in the near term helps margins, but the equity overhang looms if CleanSpark trades well above the $19.16 conversion price.
Return on invested capital plays out in two scenarios. The bull case rests on Bitcoin staying at or above $100,000, the hash price stabilizing, and the added exahashes, combined with cheap zero-percent notes, creating strong free cash flow leverage.
On the other hand, the bear case involves Bitcoin dropping or the hash price compressing further as more hashrate comes online, new capacity earns less, and dilution risk materializes with weaker equity.
The rise signals consolidation phase conditions. Cheap capital and top-quartile power costs are the main moats now. Hashpower is becoming institutionalized, with zero-percent converts, along with large Bitcoin treasuries, blurring the line between miners and structured Bitcoin funds.
CleanSpark is effectively borrowing against future mining capacity and Bitcoin holdings, treating the operation as infrastructure-backed financing rather than a speculative venture capital investment.
That’s not about survival capital. It’s the cost of entry to being structurally relevant in a one-zettahash world.
The miners who can’t access this kind of capital are getting acquired or shut down. Every dollar now has to clear a much higher hurdle than “hashrate goes up.” The narrative is tidy, and the cash flows will tell the real story.
