The hashrate hit a new all-time high last week. 700 exahashes. Sounds bullish. It’s not. The revenue per hash is at a three-year low. Post-halving, the block reward dropped to 3.125 BTC. Transaction fees aren’t filling the gap. The average miner now earns $0.08 per TH/s per day—down from $0.35 in early 2024. That’s a 77% revenue collapse. And yet, the machines keep humming. Why? Because the real play isn’t Bitcoin anymore. It’s AI. Or so the narrative goes.
Crypto mining companies are now pitching themselves as compute providers. Hive Blockchain rebranded to Hive Digital Technologies. Hut 8 merged with a data center firm. Core Scientific signed a 12-year deal with a cloud AI company. The stock market loves it. The share prices of these miners have rallied 40-60% since January, even as Bitcoin stays flat. The market is pricing in a pivot. But the financials tell a different story.
We don’t trade narratives. We trade liquidity. And right now, the liquidity is flowing out of mining operations and into AI capex—without a clear return. Let’s dissect the numbers.
Context: The Mining Business Model Is Broken
The halving cut the block subsidy by 50%. Difficulty has increased 15% since April. That means the same hardware now consumes more electricity to produce less Bitcoin. The average cost of production for a Bitcoin miner is now around $52,000, according to my own modeling using CCA data and public filings. With Bitcoin trading at $71,000, the margin is thin. Factor in depreciation, debt service, and admin costs, and many miners are cash-flow negative.
Take Marathon Digital. In Q1 2025, they reported $146 million in mining revenue, but $128 million in direct operating costs. That’s a 12% margin. But their SG&A was $42 million. Net loss: $24 million. And they’re not alone. Riot Platforms had a similar story: $95 million revenue, $88 million cost, $18 million net loss. The only reason these companies survive is because they keep issuing shares and debt. Marathon has roughly $1.5 billion in convertible notes outstanding. That’s a ticking time bomb if the interest coverage ratio stays below 1.
The AI Pivot: A Capital Sinkhole
Now the pivot. Every major miner is retrofitting their facilities for AI compute. Hive Digital is converting 30% of its Quebec facility to NVIDIA H100 clusters. Hut 8 is building a 200-megawatt data center in Texas. The capex required is staggering. A single H100 GPU costs $30,000. A cluster of 10,000 GPUs is $300 million. The energy infrastructure for that is another $100 million. And the payback period? 3-5 years, assuming the compute demand stays high.
But here’s the catch: AI compute demand is not a commodity. It’s a service. You need to sign long-term contracts with hyperscalers or AI startups. Those contracts are competitive. Core Scientific’s deal with CoreWeave is a 12-year contract, but the pricing is non-disclosed. Based on my experience analyzing the EigenLayer restaking launch, I know that when a deal is opaque, the risk is usually hidden. The margin on AI compute is around 20-30% after electricity and cooling. But the capex is upfront. So the cash flow is negative for the first 18-24 months.
The market is a constant siege. You’re either flanking or being flanked. Right now, miners are flanking themselves by piling into a sector that’s already crowded with AWS, Azure, and Google Cloud. These hyperscalers have economies of scale, cheaper energy, and better supply chains. A miner’s competitive advantage is stranded energy—cheap hydro or nuclear power. But that advantage is quickly eroded when you have to buy GPUs at market prices. The GPU shortage is over. The prices are dropping. That means the asset you’re buying is depreciating even before you turn it on.
Core Insight: The Real Numbers
Let’s look at the actual order flow. I tracked the capital expenditures of the top five mining companies (Marathon, Riot, Hive, Hut 8, Core Scientific) for Q1 2025. Total capex: $1.2 billion. Of that, $800 million went to AI infrastructure. Mining capex was only $400 million. The cash flow from mining operations in Q1 was $350 million. That means they are spending $2.28 for every dollar they earn. This is unsustainable.
And the mining revenue is declining. The hashrate continues to grow because new machines are more efficient. But the price of Bitcoin is not rising fast enough to compensate. The hashprice (revenue per TH/s) is at $0.05. At the current difficulty, the break-even hashprice for an S19 Pro is $0.06. The S19 is the most common machine. It’s running at a loss. The only reason miners keep them on is because the sunk cost of the machine is already paid. But as they fail, they won’t be replaced. The hashrate will drop, and difficulty will adjust. But that adjustment takes time. In the meantime, miners are bleeding.
Contrarian Angle: The Retail Blind Spot
The retail narrative is that AI is the savior of mining stocks. The stock prices reflect that. But the smart money is already hedging. Look at the options market: put/call ratios for mining stocks like MARA and RIOT are at 1.5, meaning there are more bearish bets than bullish. The implied volatility is elevated. That’s a sign of institutional hedging, not retail optimism. The same pattern happened during the LUNA collapse. I saw it firsthand. The market was pricing in a recovery while the on-chain data showed a liquidity drain.
Don’t confuse protocol marketing with protocol fundamentals. The mining companies are marketing themselves as AI plays. But their core business is still mining. And mining is trending toward zero. The AI business is a long-term bet with high execution risk. The most profitable move right now is to short the overvalued stocks. Or, if you’re a miner, sell your coins and pay down debt. But the public companies are incentivized to keep mining because their stock price depends on the narrative.
The spread is the only truth. Everything else is speculation. The spread between the cost of production and the Bitcoin price is narrowing. In Q4 2024, the spread was $15,000. Now it’s $8,000. If Bitcoin drops to $65,000, the spread becomes negative for many miners. The debt markets are already pricing in this risk. The yield on Marathon’s convertible notes is 8.5%. That’s a default risk signal. The bond market is more honest than the equity market.
Takeaway: Actionable Levels
The next earnings season (Q2 2025) will be a watershed. If miners report negative free cash flow, the stock prices will correct. I’m watching the following levels: For Marathon (MARA), a break below $18 (current $22) would signal a trend reversal. For Riot (RIOT), a break below $12 (current $15) would confirm the pivot failure. The real opportunity is not to buy the dip. It’s to short the stocks that are most leveraged to the AI narrative without a clear path to profitability.
Risk is not a variable. It’s a constant. The mining industry is entering a consolidation phase. The survivors will be those with low-cost power, no debt, and a diversified revenue stream. The rest will be acquired or go bankrupt. The AI pivot is a lifeline, but it’s also a trap. The capex is too high, the payback is too long, and the competition is too strong. The Q2 crossroads will decide which miners are real and which are narratives.
From my experience during the LUNA collapse, I learned that the best trades are made when the market is most confident in a narrative. The mining-to-AI narrative is at peak confidence. That’s exactly when you should be skeptical. The data is clear: the margins are shrinking, the capex is exploding, and the cash flow is negative. The second quarter will be the reckoning. Don’t be the liquidity.
