Bitcoin's network hashrate has retreated 33% from its October 2025 peak of 1.3 zettahashes per second, sliding to 861 exahashes per second.

AI is emerging as a second buyer for the assets built during the bitcoin mining boom, pushing investors to look beyond pure bitcoin production metrics. Access to power, grid interconnection and site characteristics are gaining standalone value, and that shift is starting to reshape how mining companies are valued. Keel Infrastructure, formerly Bitfarms, underscores the change. The company has exited its U.S. bitcoin mining operations and is pivoting toward high-performance computing and AI infrastructure. It cites a pipeline of up to 2.2 gigawatts of energized, secured and expansion power capacity. The significance goes beyond one strategic pivot: it signals that the infrastructure itself can be reassessed and redeployed even when bitcoin mining is no longer the primary use case. The demand driver is straightforward: AI data centers need enormous electricity supply, and new grid capacity can take years to secure. That makes existing sites with large power allocations and established interconnections valuable. Bitcoin miners spent years accumulating exactly those attributes, creating a ready-made inventory for another computing market. Riot Platforms' agreement with Anthropic provides a concrete benchmark for what AI operators may pay. The 20-year deal covers 191 megawatts at Riot's Rockdale, Texas campus and is expected to generate $9.1 billion of revenue, with extensions that could lift the total value to $16.1 billion. More than the contract itself, it highlights the arrival of an alternative market price for mining-grade power infrastructure. Still, megawatts alone do not make an AI-ready site. AI workloads typically require high-density cooling, advanced networking, upgraded electrical systems and stringent reliability standards. Converting a mining facility can require significant capital, engineering and time. As a result, physical site quality—location, permitting, connectivity, expansion potential and design constraints—may become a key differentiator, not just total power capacity. Bitcoin mining also retains an advantage AI cannot easily replicate: load flexibility. Miners can curtail when power prices rise and restart when economics improve. AI workloads usually demand consistent availability and performance. The decision is less a simple "upgrade" from mining to AI and more a tradeoff between variable upside and contracted, predictable cash flows. That optionality may become a central competitive edge. Some operators may follow Keel in moving away from bitcoin toward HPC and AI. Others may keep mining while monetizing portions of their footprint through long-term data center agreements, as Riot is doing. The common thread is that the same physical assets can increasingly serve more than one computing market. For investors, AI introduces an external reference point that the mining sector historically lacked. Mining has long been judged by bitcoin price, network difficulty, block rewards, power costs and hardware efficiency. Now an additional question is becoming relevant: what would another computing business pay for the same power and site infrastructure? The answer will vary with conversion costs, contract terms, utilization and customer concentration, but the comparison creates a market-based test for the opportunity cost of dedicating scarce infrastructure to bitcoin production. Hashrate and energy efficiency will remain core metrics, but the valuation framework is broadening. Power access, grid connectivity, site quality and the ability to support alternative workloads are moving closer to the center of how mining businesses are assessed. AI is not necessarily displacing bitcoin mining—it is creating a second market for the infrastructure behind it, and forcing the market to consider whether bitcoin will remain the highest-value customer.

Strategy CEO Phong Le said the bitcoin-treasury firm plans to resume net bitcoin purchases later this year, arguing that recent sales reflect balance-sheet management rather than a shift away from accumulation. Le said Strategy has bought roughly 175,000 BTC in 2026 while selling about 7,000 BTC, a buy-to-sell ratio of around 25-to-1. The sales have drawn attention because they depart from the company's long-running "never sell" image closely associated with chairman Michael Saylor. Strategy has sold bitcoin four times since May. The most recent disposal occurred between Aug. 3 and Aug. 9, when the company sold 1,690 BTC at an average price of $64,262, generating about $108.6 million. Strategy now holds 840,447 BTC, acquired for a total cost of about $63.36 billion at an average purchase price of $75,385 per coin, implying the latest sales were executed below its cost basis. Investors are also focused on Strategy's mNAV (multiple of net asset value), a measure comparing the firm's market capitalization with the value of its bitcoin holdings. The premium peaked near 3.4x in November 2024. By Aug. 3, Strategy's basic mNAV had slid to about 0.68x, indicating the market was valuing the company below the implied value of its bitcoin treasury. Le said proceeds from the sales have been directed toward preferred stock dividends, repurchases of its STRC preferred shares, and building a U.S. dollar reserve. Saylor, questioned on the apparent reversal, said he never promised the company would never sell, framing the phrase as personal guidance for individual holders rather than a corporate commitment. The company's Q2 2026 results added to scrutiny. Strategy reported an $8.6 billion net loss, driven largely by an $8.3 billion noncash mark-to-market loss on its bitcoin position as BTC fell toward $58,700 by quarter-end. Strategy's U.S. dollar reserve has become a key indicator of its capital-management push. The reserve rose to a record $4.65 billion in August 2026, up from $871 million in late May, more than fivefold growth in roughly ten weeks. Funding came from a mix of equity issuance and bitcoin sales, including $653.1 million raised through an at-the-market MSTR share program, plus the $108.6 million from the 1,690 BTC sale. The shift toward more flexible treasury management is not limited to Strategy. Bitcoin miner MARA sold 23,093 BTC for about $1.6 billion in the first half of 2026, citing a move away from pure accumulation. Looking ahead, Le said Strategy intends to return to net accumulation once current capital actions taper off. Traders are watching whether mNAV rebounds as buying resumes, since a prolonged discount below 1x could keep the company focused on buybacks and reserve-building rather than adding bitcoin.

Bitcoin mining is becoming increasingly subsidy-dependent as transaction fees fall back toward levels last associated with the market’s earlier era. On-chain data show fees now account for just 0.69% of miner revenue, while estimated production costs remain above spot prices and network hash rate has declined sharply from its October 2025 peak. Analysts also point to a growing shift among miners toward AI and high-performance computing (AI/HPC), raising fresh questions about longer-term incentives and network security. Key metrics - Glassnode estimates transaction fees make up 0.69% of miner revenue, near multi-year lows. Fees reportedly dipped to 0.52% in April. - Checkonchain estimates hash rate has fallen about 33% from roughly 1.3 ZH/s in October 2025 to about 861 EH/s. - Checkonchain puts the estimated average cost to produce one Bitcoin at $78,254 as of Tuesday, around 23% above the spot price cited in the source. Fees weaken, subsidy dependence grows Glassnode data indicate miners are earning little from fees relative to the fixed block subsidy. Glassnode co-founder Rafael Schultze-Kraft has said fee share stayed below 1% for almost a year and fell to 0.52% in April. In a post on X, he noted that "Bitcoin was below $400 the last time fee share was this low." The implication for miner economics is straightforward: when fee revenue compresses, operators lean more heavily on the block subsidy, which is fixed at 3.125 BTC per block. That fixed BTC payout becomes more volatile in dollar terms when Bitcoin’s price declines. The source notes Bitcoin is down nearly 50% from its October 2025 all-time high, reducing the USD value of each subsidy payment and tightening margins unless costs fall or fees rebound. Production costs exceed spot Profitability pressures are also reinforced by cost estimates. Checkonchain’s mining analytics place the estimated average cost to produce one Bitcoin at $78,254 as of Tuesday, roughly 23% above the referenced spot price. When estimated production costs outpace spot value, miners typically respond by optimizing operations, consolidating, or exiting. Smaller operators are often hit first, potentially concentrating hash rate among firms with stronger balance sheets and better power procurement. Electricity pricing and hardware efficiency remain critical inputs. With fee income subdued, any increase in non-chain costs can accelerate capacity churn. Hash rate down 33% from October 2025 peak Network-level indicators also reflect changing miner behavior. Checkonchain estimates hash rate has declined from about 1.3 ZH/s in October 2025 to roughly 861 EH/s, a drop of 33%. While hash rate alone does not define security, sustained declines can signal reduced participation and shifting incentives as difficulty and miner economics adjust. AI/HPC pivot draws scrutiny Several analysts link the hash rate decline to miners reallocating compute toward AI/HPC. Independent analyst William Clemente wrote on X that "there is no other way to slice it, hash rate has been in a decline," citing margin compression since 2022 and higher energy prices, while also highlighting that many operators are pivoting into "more lucrative AI computing." He described AI-related shifts as "prudent business decisions" for publicly traded companies. The concern raised by analysts is not diversification itself, but the timing: mining participation may be slipping while fee revenue remains under 1% of miner income and the subsidy’s USD value is pressured by price weakness. The source cites prior reporting that CleanSpark refocused on AI by shifting toward operating data centers after missing profit targets. It also references Keel Infrastructure, which shut down all U.S. mining operations after revenue fell 50% in the second quarter. Capriole Investments founder Charles Edwards also tied hash rate declines to the AI pivot, calling it a "concerning Bitcoin development in 2026" and saying the trend accelerated since April. What investors are watching With fee revenue near decade lows and estimated mining costs above spot, the next key signal is whether hash rate stabilizes as difficulty adjusts or whether more operators continue shifting capital and compute into AI/HPC. Markets will also be watching for any sustained recovery in fee share, since prolonged low fees increase reliance on the subsidy at a time when mining economics and miner strategy appear to be in flux.

On-chain tracker Lookonchain reported that as BTC climbed, wallet address 0xff84 came close to liquidation on a 1,793 BTC short position worth about $114.4 million. The address later reduced exposure, leaving a remaining short of 1,543 BTC, or roughly $98.97 million. The latest liquidation price has moved up to $64,225.35.

Strategy has made its Bitcoin Treasury framework more transparent by releasing a public Bitcoin Credit Model, a live dashboard that breaks down the credit profile of its $53.5B BTC Treasury. The tool tracks credit spreads and undercollateralization risk across $21.95B of debt and preferred stock, using a 10% BTC annualized rate of return (ARR) reference case to gauge solvency. It also assigns each instrument a color-coded status—Investment Grade, HighYield, or Distressed—and calculates the BTC floor price needed to keep the stack fully collateralized.

Publicly listed bitcoin miners have stepped up coin sales this year, adding a steady source of supply to an already soft market. CoinDesk cited Blockware Intelligence data showing these miners held about 127,000 BTC at the start of 2026. That figure has fallen to roughly 99,000 BTC, implying around 28,000 BTC sold year to date—about $1.78 billion at current prices. While the volume is smaller than the more than $4.4 billion of net outflows seen from U.S. spot bitcoin ETFs, analysts say persistent, incremental selling can weigh more than many investors expect when price trends are down and spot demand is thin. Bitcoin is down 27% since the start of 2026, lagging major assets including the S&P 500. Mining economics have also tightened, with average production cost estimated near $74,300 per BTC, pushing more operators to prioritize AI computing infrastructure. On the network side, mining difficulty has dropped about 18% from its November peak, lifting mining rewards for remaining miners by roughly 18% and altering competitive dynamics across the sector.

On-chain tracking service Lookonchain (@lookonchain) said that as BTC climbed, address 0xff84's 1,793 BTC short—worth about $114.4 million—moved to the brink of liquidation. The wallet later reduced exposure by closing part of the position. It now holds a remaining short of 1,543 BTC, or roughly $98.97 million, with the updated liquidation price lifted to $64,225.35.

Odaily Planet Daily cited Lookonchain data showing that Bitcoin's latest move higher has pushed a large short position close to liquidation. The wallet previously held a 1,793 BTC short worth $114.4 million and has already reduced exposure to avoid being liquidated. It now remains short 1,543 BTC, valued at $98.97 million, with the liquidation price reset to $64,225.35.