MARA Stock Deep Dive: Why Did Mara Holdings Shift from Bitcoin Mining to AI?

In the first quarter of 2026, Bitcoin mining company MARA Holdings completed a thorough strategic shift. The company sold approximately 15,133 Bitcoins in batches from March 4 to 25, realizing about $1.1 billion, reducing its holdings from 53,822 to 38,689 Bitcoins, a decrease of roughly 28%. If the entire quarter is included in the calculation, MARA sold 20,880 Bitcoins at an average price of $70,137 per Bitcoin, totaling approximately $1.5 billion. Alongside this large-scale sell-off, the company explicitly announced that it would no longer make large-scale purchases of ASIC miners in the future, shifting its focus toward artificial intelligence and high-performance computing infrastructure.

This strategic pivot was directly driven by financial pressures. Bitcoin prices in the first quarter of 2026 fell from about $87,000 to approximately $68,000, a decline of about 22%, which directly caused a significant decrease in the fair value of MARA’s Bitcoin holdings on its balance sheet. The company reported a net loss of about $1.3 billion, with unrealized fair value losses on digital assets reaching as high as $1 billion. Meanwhile, mining output during the quarter was only 2,247 Bitcoins, with an average production cost of $76,288, while the market price at the time of sale had already dropped to $70,137, indicating that the mining business itself was already showing accounting losses.

Beyond financial statements, MARA is attempting to address more fundamental structural issues: the business model relying solely on Bitcoin mining is rapidly losing resilience under the dual shocks of halving cycles deepening and rising energy costs. The shift toward AI infrastructure is a strategic choice made by this once second-largest Bitcoin reserve holder in the context of industry reshaping.

How do the contradictions between financial losses and hash rate expansion occur simultaneously?

In the first quarter of 2026, MARA set a record with an energized hash rate reaching 72.2 EH/s, a 33% increase year-over-year. However, this operational expansion did not translate into financial gains. During the same period, the company's revenue was $174.6 million, down 18% year-over-year, falling short of analyst expectations of $192.7 million. The core reason for revenue decline was not a decrease in output—hash rate growth actually partially offset production decline—but rather an approximately 18% drop in the average selling price of Bitcoin compared to the same period last year, which significantly shrank the fiat value generated per block.

On the cost side, MARA’s electricity procurement costs rose from $35,728 per Bitcoin last year to $40,047. When factoring in depreciation, management, and debt costs, the total cost per Bitcoin is estimated to be between $60,000 and $70,000. This means that the low point of Bitcoin prices in the first quarter was near or even touching the breakeven line.

Another major source of increased losses is the impact of accounting standards. MARA’s Bitcoin holdings are measured at fair value, so a price decline directly impacts current earnings, resulting in accounting losses. These are unrealized losses and do not involve cash outflows, but they reveal the balance sheet risk of holding assets during a market downturn. After selling Bitcoin at a discount to repurchase convertible notes, MARA’s total debt decreased from $3.3 billion to about $2.3 billion, reducing by roughly $1 billion, while saving approximately $88.1 million in interest expenses.

How has MARA’s position shifted within the global mining industry’s competitive landscape?

In terms of hash rate, MARA still belongs to the top tier of the industry. However, its market capitalization ranking has fallen from the largest Bitcoin miner to seventh place. This change is not entirely due to MARA’s own performance decline—some competitors have gained more market recognition through progress in AI transformation and capital structure optimization, making their valuations more resilient.

In terms of listed company Bitcoin holdings, MARA’s ranking has been more volatile. After the first quarter’s sell-off, the company dropped from second to fourth place among publicly listed companies holding Bitcoin. Strategy (formerly MicroStrategy) remains first with about 843,738 BTC holdings, while Japan’s Metaplanet jumped to third place due to frequent increases in holdings during the first quarter of 2026. The rise and fall in MARA’s ranking reflect a profound divergence in the strategies of listed companies’ Bitcoin reserves: Strategy has built ongoing buying capacity through preferred stock financing, whereas MARA has chosen to allocate funds toward new business directions outside of hash rate.

How does the acquisition of Long Ridge support MARA’s transition to AI computing infrastructure?

On April 30, 2026, MARA announced the acquisition of Long Ridge Energy & Power in Ohio for approximately $1.5 billion, including a 505 MW combined-cycle natural gas power plant and over 1,600 acres of contiguous land. The plant achieved an annualized adjusted EBITDA of about $144 million in the second half of 2025, with approximately 76% of its capacity secured through swap agreements at a weighted average price of $38.80 per MWh, locking in revenue. Its total operating costs are below $15 per MWh. The facility also supplies about 100 million cubic feet of associated natural gas daily, providing over 20 years of self-supplied fuel.

This acquisition increased MARA’s owned and operational capacity by 65%, reaching approximately 2.2 GW across PJM, ERCOT, SPP, and international markets. According to plans, Long Ridge can be scaled up to about 600 MW of AI load in the short term, with a total potential capacity exceeding 1 GW. The initial phase of AI and critical IT load construction is expected to start in the first half of 2027, with the first capacity reaching commercial operation by mid-2028.

In addition to Long Ridge, MARA also established a strategic partnership with Starwood Digital Ventures in February 2026, planning to convert some existing mining farms into AI data centers. The goal is to provide over 1 GW of immediate IT capacity, with MARA retaining up to 50% equity in the joint venture to participate in future cash flows. The company also acquired a 64% stake in Exaion to strengthen capabilities in enterprise AI and sovereign cloud infrastructure.

From a strategic perspective, MARA’s core idea is shifting from “electricity used for mining” to “owning power and allocating it to the highest-value applications”—including AI training and inference, critical IT loads, and Bitcoin mining. About 90% of its self-operated mining capacity could be reallocated to AI and IT operations in the future. This approach of viewing electricity resources as core assets and entering the AI computing market is a distinctive feature among U.S. mining companies.

What are the core disagreements in the market regarding the sale of Bitcoin and the shift to AI?

MARA’s large-scale sell-off and strategic shift have created significant divergence within the market.

Bearish view: argues that MARA sold a large amount of Bitcoin at an average price of $70,137 near the market bottom in March 2026, after just a few months having raised funds through issuance of zero-coupon convertible notes to increase Bitcoin holdings. This behavior of switching from “buy and hold” to selling for financing in the short term may indicate that the company’s confidence in its Bitcoin reserve accumulation model has substantially waned. Although the AI transition offers a new narrative, this field involves intensive investment and fierce competition. MARA has not yet signed any HPC lease agreements, so the validation of its AI business model lacks key catalysts.

Bullish view: believes that MARA’s transition decisions are based on quantifiable asset conditions. Its owned sites have electricity costs of about $0.04 per kWh, among the lowest in the U.S. mining industry. Using this as a foundation, directing power resources toward AI infrastructure is a re-pricing of existing assets rather than abandoning the original business. Debt buyback at a discount has generated about $71M in accounting gains and eliminated ongoing interest expenses, which is meaningful for capital structure optimization. From a long-term asset allocation perspective, owning power plants and data centers provides more stable cash flows and diversified monetization paths compared to simply operating mining equipment.

How do regulatory policies in the U.S. influence MARA and the long-term outlook of the American Bitcoin mining industry?

By 2026, the U.S. federal regulatory landscape for Bitcoin mining shows a complex dual signal.

On the supportive side, Republican senators announced the “Mined in America Act” in March 2026, proposing a federal voluntary certification program to support domestic cryptocurrency mining and gradually phase out reliance on foreign-manufactured hardware. Additionally, the SEC and CFTC issued joint guidelines clarifying that Bitcoin proof-of-work mining rewards are not considered securities, removing a long-standing regulatory uncertainty in the field.

On the restrictive side, the Biden administration reintroduced a proposal for a “Digital Asset Mining Energy Tax” in March 2026, aiming to impose a 30% excise tax on the electricity costs used for Bitcoin mining. If enacted, this tax would directly erode the cost advantages of U.S. miners including MARA.

The tension between these two policy directions constitutes a key variable for the industry’s long-term development. Large-scale enterprises with stable power supplies and compliant operations are better equipped to withstand the impact of energy taxes and meet certification requirements compared to dispersed small and medium miners. MARA’s shift from “electricity consumer” to “power owner” can also be seen as a preemptive response to potential policy risks: controlling power generation capacity allows the company to achieve cost adjustments through vertical integration and reduce sensitivity to external electricity market pricing and tax policy changes.

Is the “hash rate migration” among Bitcoin miners a structural upgrade or a strategic retreat?

MARA’s strategic shift is not an isolated phenomenon. In the first quarter of 2026, several Nasdaq-listed miners began transformation almost simultaneously. Core Scientific’s AI hosting revenue has surpassed Bitcoin mining income, becoming its largest revenue source. IREN signed a five-year, $3.4 billion AI cloud service contract with NVIDIA, with AI cloud revenue increasing by 94.2% quarter-over-quarter. By the end of Q1, the total AI and HPC-related contracts signed by listed miners exceeded $70 billion.

The underlying driver of this shift is the structural deterioration of unit economic efficiency in mining. The halving of Bitcoin’s block reward from 6.25 BTC to 3.125 BTC has been over a year, and the halving effect is fully reflected in the revenue structure. Meanwhile, total network hash rate continues to climb, while Bitcoin prices have yet to fully recover. The cost per Bitcoin has approached or even exceeded market prices at times, and the marginal returns from simply expanding hash rate are diminishing.

MARA’s approach differs from peers: Core Scientific and IREN rely more on third-party contracts to provide hosting services for AI computing power, while MARA emphasizes deep vertical integration by directly owning power generation assets. Which model can maintain a long-term advantage in future market competition depends on a comprehensive comparison of capital efficiency, contract pricing ability, and operational execution.

Summary

MARA Holdings’ strategic shift in the first quarter of 2026 is a sample-level structural adjustment in Bitcoin mining amid the deepening halving cycle and explosive demand for AI computing power. The company completed deleveraging and capital restructuring through the sale of about $1.5 billion worth of Bitcoin, shifting resources from traditional hash rate expansion to AI and HPC infrastructure. Although short-term financial data are under significant pressure due to fair value changes in Bitcoin, its core assets—electricity resources—and the acquisition of Long Ridge Power Plant have laid a foundation for transitioning toward a digital infrastructure platform.

Current market assessments of MARA’s transformation focus less on whether the AI narrative is valid, and more on whether the timeline for translating this narrative into actual cash flows can match the patience of capital markets. The first phase of AI load construction at Long Ridge is not scheduled to begin operation until mid-2028. During this period, the company still needs to rely mainly on Bitcoin mining revenue to sustain operations and support large-scale capital expenditures. Whether MARA will become the next successful case of power asset monetization after CoreWeave, or face cash flow pressures and financing challenges amid intense capital competition, remains to be seen from subsequent financial quarters’ operating data.

FAQ

Q: How many Bitcoins did MARA sell in the first quarter of 2026, and what was the reason?

MARA sold approximately 20,880 Bitcoins at an average price of $70,137 per Bitcoin in the first quarter of 2026, totaling about $1.5 billion. The proceeds were mainly used to repurchase convertible bonds at a discount, reducing total debt from $3.3 billion to $2.3 billion, and to fund expansion into artificial intelligence and data center infrastructure.

Q: What are MARA’s current hash rate and Bitcoin holdings?

As of the end of the first quarter of 2026, MARA’s energized hash rate reached a record 72.2 EH/s, up 33% year-over-year. After the sell-off, the company’s Bitcoin holdings decreased from 38,689 to approximately 35,303 Bitcoins.

Q: What is the core asset in MARA’s transition to AI infrastructure?

On April 30, 2026, MARA announced the acquisition of Ohio’s Long Ridge Energy & Power facility for about $1.5 billion, including a 505 MW natural gas power plant and over 1,600 acres of land, capable of supporting over 600 MW of AI load in the long term. The company also partnered with Starwood Digital Ventures to convert existing mining farms into AI data centers.

Q: How has MARA’s ranking in listed company Bitcoin holdings changed?

After the first quarter’s Bitcoin sell-off, MARA’s ranking among publicly listed companies holding Bitcoin dropped from second to fourth place.

Q: What are the key risks MARA faces in its AI transition?

MARA has not yet signed any HPC lease agreements, and the validation of its AI business model lacks key catalysts. The first phase of AI load construction is not expected to start until mid-2027, with commercial operation targeted for mid-2028. During this period, the company may face significant capital expenditure pressures and market doubts about revenue visibility. Additionally, ongoing Bitcoin price volatility continues to directly impact MARA’s mining profitability.

Q: How do U.S. regulatory policies affect the Bitcoin mining industry?

In 2026, the U.S. federal regulatory landscape shows divergence: on one side, the “Mined in America Act” supports domestic mining and establishes a voluntary certification system; on the other, the Biden administration proposed a 30% excise tax on electricity used for Bitcoin mining. Meanwhile, the SEC and CFTC clarified that Bitcoin proof-of-work rewards are not securities, removing a key regulatory uncertainty.