BitPlanet Says Slumping Bitcoin Mining Profitability Is Driving Miners Toward AI, HPC

BitPlanet Research Lab said Sept. 2 that it had published a report titled "The Economics of Mining in 2026 Reshaped by Record-Low Hashprice."

The report said structural pressure continues to build across the Bitcoin mining industry as profitability worsens. All three of the lowest monthly average hashprice readings on record were posted this year, it said, cutting into miners' revenue.

Monthly mining revenue fell 23% to $836.41 million in June from $1.086 billion in May. Revenue rebounded to $875.35 million in July, but that recovered only about 16% of June's decline. Of the 17 mining difficulty adjustments made this year through Aug. 22, 10 were decreases.

Major miners are also responding to the squeeze. The report said Marathon Digital Holdings, the world's largest mining company by scale, sold about 91% of the Bitcoin it mined in the second quarter, while CleanSpark disposed of a substantial portion of the Bitcoin it mined in July.

Miners are also diversifying into artificial intelligence and high-performance computing rather than relying only on mining. The cumulative value of AI and HPC contracts announced by publicly traded miners exceeded $70 billion as of the first quarter, the report said. As mining profitability declines, companies are seeking to use their power capacity and land for higher-return businesses such as AI and HPC.

Still, whether a shift into AI and HPC will improve profitability remains unclear. Building mining infrastructure is estimated to cost $700,000 to $1 million per megawatt, compared with $8 million to $15 million for AI infrastructure, requiring far more upfront investment.

BitPlanet said the current shift is less a contraction of the mining industry as a whole than a divergence in strategy based on the economics of individual operators. Companies with access to low-cost power and high-efficiency equipment can continue mining, while operators without those advantages have a growing incentive to convert facilities to other uses such as AI and HPC.