Bitcoin miners shift from crypto to high-performance computing

Bitcoin climbed 21.5% between the close on Aug. 17 and Aug. 21, yet six out of seven major US-listed mining companies finished that same trading period significantly lower. While MARA Holdings advanced 16.1% to track BTC most closely, Cipher Digital dropped 14.8%, TeraWulf lost 11.2%, Hut 8 fell 8.1%, and IREN decreased by 6.8%.

These equities, historically traded as pure Bitcoin proxies, diverged during one of the cryptocurrency’s strongest weekly performances of the year.

During the same sessions, the Invesco QQQ Trust (QQQ) dropped 2.3% amid fluctuating long-term yields, pulling the miners into a softer technology-equity environment. This shift is partly driven by their corporate structures, as several legacy mining firms now generate revenue, secure financing, or derive forward valuations from long-term data-center agreements.

Because identical electricity supplies, land parcels, and grid connections can power either ASIC miners or GPU clusters, public markets evaluate each application through distinct risk frameworks.

CryptoSlate reviewed two years of daily closing prices to determine whether the August divergence formed part of a broader trend. While sensitivity to Bitcoin has declined across most of the sector as data-center agreements take on greater significance, the rate coefficients vary too widely to classify every AI-focused miner strictly as a long-bond proxy.

Power is the second product for Bitcoin miners

Traditionally, a mining operator uses specialized hardware to convert electrical power into Bitcoin, making its financial performance reliant on coin prices, network difficulty, transaction fees, fleet efficiency, and electricity costs.

High fixed expenses amplify this dynamic, allowing percentage gains in Bitcoin to yield even larger percentage increases in expected equity value when revenues outpace baseline costs.

Holding mined Bitcoin on the balance sheet introduces an additional layer of market exposure, particularly when companies fund expansion while retaining the bulk of their output. Historically, investors viewed mining stocks as leveraged Bitcoin investments burdened with corporate, financing, and operational risks.

This dynamic functioned effectively when mining constituted nearly all revenue and executive teams focused capital allocation entirely on hash rates.

The emergence of artificial intelligence infrastructure disrupted this model, transforming electricity into a contested resource sought by both sectors. Miners equipped with grid interconnection agreements can lease power capacity to hyperscalers or establish GPU cloud operations, trading unpredictable mining returns for credit-backed tenant contracts.

To fund these expansions, firms are utilizing debt financing and liquidating Bitcoin reserves, introducing construction timelines, equipment procurement hurdles, and customer concentration risks to valuations that once relied solely on hash prices.

Corporate filings reveal that these publicly traded miners are at varying stages of transitioning toward AI. TeraWulf reported $31.9 million of its $44.8 million second-quarter revenue from high-performance computing leases, alongside roughly $12.8 million from digital assets.

Hut 8 reported that its Beacon Point leases span 949 megawatts of contracted IT capacity, representing $26.6 billion in base-term contract value, pending future deployment and tenant performance.

IREN posted $70.5 million in AI cloud revenue compared to $66.7 million from Bitcoin mining during its June quarter, pushing AI ahead of mining in its current revenue breakdown. An Aug. 27 announcement also reported an operating annual recurring revenue (ARR) of $1 billion as of Aug. 26, alongside $4 billion in contracted ARR tied to 2026 capacity—scheduled to become operational by Dec. 31, contingent on commissioning, testing, and client approval.

That same quarter included a $450.4 million impairment largely associated with decommissioned mining equipment as facilities pivot toward AI operations.

Cipher continued to generate revenue from Bitcoin mining in the second quarter, but has secured 700 megawatts of high-performance computing capacity across three locations, initiating initial capacity deliveries at its Black Pearl site in August.

Riot Platforms occupies a middle ground, reporting $113.7 million in mining income, $23.2 million from data centers, and $37.3 million from engineering services within a $174.2 million quarterly total. Its 241 megawatts of contracted AI capacity represent approximately $9.8 billion in estimated long-term revenue, offering investors a contracted order book to evaluate alongside the 11,380 Bitcoin held on June 30.

CleanSpark operated as a dedicated mining enterprise earlier in the year, though that designation shifted prior to the August rally. The company executed a 20-year, $6.6 billion data-center lease on Aug. 6 while continuing to generate operating revenue exclusively from mining, integrating the firm into the newly hybrid cohort.

Meanwhile, MARA functions as the closest major mining-centric benchmark, even as it investigates auxiliary energy and computing ventures.

The split shows up in the beta

CryptoSlate analyzed historical Alpaca/IEX equity closing prices for HUT, WULF, IREN, CIFR, RIOT, MARA, CLSK, and QQQ, alongside Alpaca’s BTC/USD closing prices, spanning from Aug. 22, 2024, to Aug. 24, 2026.

Daily equity returns were tracked across market trading days, with each Bitcoin return measuring the span between consecutive stock market sessions, thereby incorporating weekend movements into Monday data. Ten-year Treasury yields were sourced from the Federal Reserve’s DGS10 data series.

Close-to-close returns from Aug. 17 through Aug. 21 illustrate the initial market separation by evaluating Bitcoin’s upward movement alongside QQQ as a baseline reference for the broader technology sector during the same timeframe.

The extended evaluation utilized rolling 90-trading-day correlations and univariate Bitcoin betas, contrasting a period concluding on Aug. 22, 2025, with a current window ending Aug. 24, 2026.

A Bitcoin beta of 1.10 indicates that a 1% daily shift in Bitcoin historically corresponded to an average 1.10% movement in the miners over that timeframe, whereas correlation gauges the consistency of their directional alignment and magnitude.

Bitcoin betas decreased compared to the 2025 benchmark window for six of the firms—with IREN remaining stable at 0.93—while Bitcoin correlations declined for six companies and ticked upward for WULF from an already low 0.17 to 0.22.

Correlation with the QQQ surpassed Bitcoin correlation across all seven entities in the current dataset, demonstrating that their daily equity performance tracked the Nasdaq-backed index more consistently than the underlying cryptocurrency.

MARA maintained the highest Bitcoin correlation and beta within the group, reflecting its persistent reliance on traditional mining economics.

HUT, WULF, and CIFR populate the lower tier of current Bitcoin correlations as data-center agreements assume greater prominence in valuation metrics, whereas MARA remains at the upper extreme. IREN diverges from this simple hierarchy; its Bitcoin beta remained steady even as its QQQ correlation climbed to 0.60, blending an active mining operation with an extensive AI infrastructure pipeline within a single return series.

CryptoSlate also executed a multi-factor regression analysis for year-to-date 2026 incorporating daily Bitcoin returns, QQQ returns, and daily shifts in 10-year Treasury yields. This three-factor model accounted for roughly 28% to 45% of daily price variation across the seven firms, with a projected 10-basis-point yield increase corresponding to outcomes ranging from a 0.52% drop for WULF to a 0.79% rise for CIFR.

Because four rate coefficients registered as negative and three as positive, the dataset does not substantiate a unified duration-trading pattern across the sector.

Conflicting coefficients complicate simple bond analogies, as higher yields diminish the present value of long-term cash flows and increase financing expenses, while daily equity prices simultaneously reflect corporate announcements, infrastructure updates, treasury holdings, energy tariffs, and capital raises.

While contract details account for differing exposures across businesses, the metrics fail to indicate a shared interest-rate trading strategy for the group.

Contracts give the miners a second risk map

TeraWulf serves as a primary example of a firm whose financial statements have transitioned into a new operational category. High-performance computing leases generated roughly 71% of its second-quarter revenue, and corporate disclosures outline plans to repurpose or curtail mining hardware as compute capacity scales up.

While a Bitcoin rally can enhance residual mining yields, equity valuations now hinge equally on client disbursements, construction milestones, and campus-level financing structures.

Cipher illustrates how market perception can shift ahead of shifts in revenue composition. Its June quarter continued to reflect mining output, yet investors are able to price in contracted compute capacity alongside an estimated $793 million in average annual net operating income tied to its baseline lease terms.

The discrepancy between current earnings and anticipated capacity makes delivery schedules, capital expenditure, and counterparty creditworthiness vital components for share valuations.

Riot and IREN display both exposures more visibly, despite their divergent revenue distributions. Mining accounts for the majority of Riot’s current income, whereas IREN’s June results positioned AI cloud earnings slightly ahead of digital asset mining.

Their underlying mining operations retain sensitivity to asset pricing, while formalized AI contracts introduce additional streams of projected cash flow. Because each quarterly report alters the relative importance of Bitcoin production versus data-center buildouts, beta functions as a reflection of business mix rather than a fixed corporate attribute.

Announcements of enterprise contracts carry inherent uncertainties, as headline figures denote multi-year expected payouts, projected revenue and net operating income depend on adherence to deployment timelines, and project-specific debt insulates parent entities only according to legal structuring.

Multi-year aggregate valuations signal corporate direction, but they cannot replace traditional discounted cash-flow models or physical asset commissioning.

The Aug. 17 market breakout captured a genuine structural divergence, and the extended dataset reinforces a broader industry reclassification. MARA advanced alongside the cryptocurrency, every firm maintained a positive contemporary Bitcoin beta, and legacy mining cash flows continue to support various AI infrastructure expansions.

Bitcoin has transitioned into one component among many, with its influence diminishing wherever contracted computing infrastructure dictates the core investment thesis.

The term “Bitcoin miners” now serves more as a historical descriptor than an indicator of corporate destination. Investors purchasing the sector via a mining index obtain exposure to Bitcoin production, hyperscaler credit risk, construction timelines, power distribution liabilities, project financing, and technology equity multiples in varying degrees.

While contract agreements account for why these equities decoupled, residual Bitcoin betas demonstrate that their legacy identities continue to influence their market performance.