Bitcoin’s Hashrate Bear Market: Why Miners Are Pivoting to AI

Bitcoin mining is facing a change that is larger than a temporary shutdown or a routine market pullback. After reaching a late-2025 high, the network’s total hashrate—the computing power used to mine Bitcoin—entered what Twenty One Capital CEO Rapha Zagury described as the network’s first “hashrate bear market.” The term is an industry description, not an official Bitcoin protocol classification. 12

The central tension is straightforward: Bitcoin mining revenue is volatile, while AI and high-performance computing (HPC) operators are willing to sign long-term contracts for access to power, land and data-center capacity. As a result, some miners are reallocating infrastructure toward AI workloads rather than continuing to expand Bitcoin mining at any cost.

How far has Bitcoin’s hashrate fallen?

Bitcoin’s 30-day average hashrate reportedly declined from about 1,108 exahashes per second (EH/s) in November 2025 to 898 EH/s by August 2026, a fall of roughly 19%. The decline followed a network peak near the end of 2025 and represented an unusually sustained contraction rather than a short-lived outage. 24

Other measurements show a more uneven path, including a partial recovery after an early-2026 low. CoinShares reported that hashrate fell from its October peak to about 850 EH/s by early February before recovering, while first-quarter hashrate still ended below where it began. 5910 These figures use different periods and measurement windows, so they should not be treated as a single real-time reading. They do, however, point to the same broad conclusion: Bitcoin’s long-running hashrate growth has temporarily reversed.

Why miners are moving capacity to AI and HPC

Bitcoin miners have several assets that AI data-center operators need:

• access to large power allocations;
• land and grid interconnections;
• cooling and electrical infrastructure; and
• experience operating energy-intensive facilities.

Those assets can often be repurposed for AI or HPC hosting, although the transition requires significant capital. One report cited estimated conversion costs of roughly $8 million to $15 million per megawatt, including cooling systems. 3

The economic comparison is the main incentive. Bitcoin mining revenue per unit of computing power changes with the price of BTC, transaction fees, network difficulty and electricity costs. AI hosting, by contrast, can offer contracted revenue over a defined period. Reports indicate that rates for leasing power facilities to AI operators have exceeded Bitcoin’s hashprice in many cases. 3

This is not a conversion of Bitcoin ASIC miners into AI processors. It is a shift in how companies use their sites, power capacity and data-center infrastructure.

The contract boom behind the pivot

CoinShares counted more than $70 billion in announced AI/HPC contracts across listed Bitcoin miners. 9 The most prominent example is Core Scientific’s relationship with CoreWeave.

The CoreWeave arrangement expanded to approximately $10.2 billion over 12 years. CoinShares reported that about 350 MW had been energized for HPC, with a longer-term target of roughly 590 MW. 9 Core Scientific also added about 70 MW under its long-term relationship with CoreWeave. 16

The significance of these contracts is not just their headline value. They show why a mining company might prioritize infrastructure leasing over adding more Bitcoin machines: a long-term customer can make a power-intensive site easier to finance and may reduce exposure to daily swings in mining revenue.

Falling hashprice is squeezing mining margins

Hashprice measures the daily revenue generated by a unit of Bitcoin mining capacity. CoinShares-linked reporting put hashprice at roughly $36–$38 per petahash per day in the fourth quarter of 2025, before it fell to around $29 in the first quarter of 2026. 714

At the same time, the average cost for listed miners to produce one Bitcoin was estimated at about $80,000 in Q4 2025. 14 That does not mean every miner faced the same cost: electricity contracts, machine efficiency, financing, curtailment arrangements and accounting methods differ significantly. It does show how quickly margins can become thin when BTC prices weaken while power and infrastructure costs remain fixed.

The pressure has encouraged miners to find revenue outside block rewards. Publicly listed miners reportedly sold more than 32,000 BTC in Q1 2026, using treasury assets for debt repayment, liquidity and other corporate needs. 12

What MARA’s Bitcoin sale reveals

MARA Holdings provided one of the clearest examples of the industry’s financial pressure. The company announced that it sold 15,133 BTC for approximately $1.1 billion between March 4 and March 25, 2026. It used the proceeds to repurchase approximately $1 billion of 0% convertible notes due in 2030 and 2031. 2021

The transaction does not prove that MARA has stopped mining. It does show a shift away from automatically retaining all mined Bitcoin. A miner can use its treasury as a source of liquidity, reduce debt or fund a broader infrastructure strategy instead of treating BTC accumulation as the only route to growth.

That distinction matters. Treasury sales, AI hosting and continuing mining operations can all occur at the same company. The sector’s pivot is better understood as diversification and reallocation than as a clean overnight exit from Bitcoin mining.

Which miners are still operating?

Major operators including MARA, Riot, CleanSpark, Core Scientific, Hut 8, Cipher Mining, IREN, Bitfarms and Cango continue to mine or retain substantial mining operations, while several are also developing AI/HPC businesses. 1216

“Still operating” does not mean “unaffected.” Companies can reduce active hashrate, redirect individual sites, sell BTC reserves or prioritize AI capacity without abandoning Bitcoin entirely. The balance between mining and AI hosting can also change as BTC prices, electricity contracts and customer demand change.

The emerging divide is therefore less about miners versus non-miners and more about business models:

• Pure-play miners remain more directly exposed to BTC price, network difficulty and electricity costs.
• Infrastructure companies can potentially combine mining revenue with contracted AI/HPC income.
• Highly leveraged operators may need to sell BTC or other assets to manage debt during periods of weak margins.

Why lower hashrate can help the miners that remain

Bitcoin’s difficulty adjustment is designed to respond to changes in the amount of mining power online. If enough machines disconnect, mining difficulty eventually falls. The miners that remain online can then receive a larger expected share of the network’s block subsidy and transaction fees for each unit of hashrate they operate. Reported difficulty fell about 19.9% from its cited peak during the decline. 4

That mechanism can improve the relative position of efficient survivors. A miner with low-cost electricity and newer ASICs may become more competitive when less efficient machines leave the network.

But difficulty adjustment does not create new Bitcoin revenue. It redistributes the existing reward opportunity among fewer machines. Profitability still depends on:

• Bitcoin’s market price and transaction fees;
• electricity, cooling and curtailment costs;
• ASIC efficiency and replacement costs;
• financing and other capital expenses; and
• how much attractive power capacity is redirected to AI and HPC.

What happens next?

The AI pivot could strengthen diversified miners by giving them contracted revenue and a second use for their power infrastructure. At the same time, a deeper shift could reduce the hashrate supporting Bitcoin and make the network’s mining sector more concentrated among the most efficient operators.

The trend is not irreversible. A sustained BTC rally, cheaper electricity or weaker AI-hosting demand could make Bitcoin mining more attractive again. For now, however, the evidence points to a structural contest for power and data-center capacity—not simply a conventional Bitcoin mining downturn. The companies best positioned to survive may be those that can remain efficient Bitcoin miners while monetizing infrastructure in a market where AI customers are willing to pay more for scarce electricity and facilities.