The market is witnessing a tectonic shift in the strategy of public bitcoin miners. In the first six months of the year, the combined realized hashrate of these companies fell by 56 EH/s, equivalent to a 15% decline. However, this is not a classic exit from the industry, but rather a clever restructuring: a significant portion of capacity has not been liquidated, but repurposed for high-performance computing (HPC) and artificial intelligence.
The financial arithmetic of the transition
An analysis of second-quarter reports reveals explosive growth in revenues unconventional for miners. Total disclosed revenue from HPC and AI segments jumped 52% compared to the previous quarter. Notably, about 10% of capacity was literally redirected from mining digital gold to servicing AI workloads. For some market players, revenue from colocation and cloud services exceeded proceeds from winding-down mining for the first time in history.
Nevertheless, the price of this technological rearmament is steep. Total costs for infrastructure transformation exceeded $30 billion. Particularly telling is the gap between capital expenditures and current revenue: among six major infrastructure providers with recurring HPC revenue, capital expenditures were nearly 15 times their total revenue for the reporting period. This is a classic story of "heavy investments" with deferred profit.
Economics per megawatt-hour
The performance figures look intriguing. The revenue per MWh metric among six HPC service providers ranges from $86 to $300, with a median value of about $180. For comparison, AI cloud services show an average estimate of $940.74 per MWh—a multiple gap compared to traditional mining. For reference: Bitmain's flagship ASIC miner, the Antminer S23, generates only about $179.13 per MWh, while the older S21 Pro model shows a modest $113.45.
It is important to understand the fundamental difference in business models. HPC colocation involves long-term contracts with predictable cash flow, often allowing the burden of energy costs to be shifted onto the client. Bitcoin mining, on the other hand, remains a high-risk endeavor tied to the volatility of the asset's price, network difficulty, and transaction fees. Interestingly, even altcoins sometimes look more attractive: equipment for mining Zcash (Z15 Pro) temporarily brings in about $585.61 per MWh, outpacing HPC colocation, albeit with a higher degree of risk.
The situation with market leaders is also telling. The largest American company, MARA Holdings, ended the second quarter with a net loss of $611.3 million, whereas a year earlier it recorded a profit of $808.2 million. This clearly demonstrates that the era of the "money printer" for miners is over, and now only those capable of adapting to the new reality will survive.
My view: The current environment is not capitulation, but evolution. Miners are transforming into versatile energy operators, hedging cryptocurrency market risks with stable AI contracts. However, the colossal gap between CAPEX and OPEX at this stage makes such companies vulnerable to debt pressure. The success of this strategy will depend on management's ability not just to fill capacity, but to secure long-term contracts with adequate margins; otherwise, we risk seeing a wave of consolidation as early as the next cycle.