MARA Expands Planned Power Portfolio to 4.2GW; First AI Contract Could Be Key to Re-rating

Overview

After MARA Holdings released its 2Q26 results, Bernstein focused on assessing its progress in extending from Bitcoin mining into AI and high-performance computing infrastructure. The company is expanding controllable power resources through acquisitions and partnerships, and is advancing lease negotiations at multiple sites with Starwood's engineering, procurement, construction, and development capabilities. The report believes that the strategic direction has made substantive progress, but the market is still waiting for the first commercial AI contract to validate demand, economics, and execution capability.

Core views

The core view is that MARA now has a fairly substantial base of power resources and multiple pathways for AI deployment, but has not yet achieved decisive commercial validation. The Matagorda County and Long Ridge projects could significantly expand planned capacity, while the Starwood partnership should help reduce the development execution burden and improve capital efficiency. Management reiterated its goal of signing at least two AI agreements by year-end; if contracts are secured, they could drive a valuation re-rating. Until then, Bernstein maintains a neutral stance.

Analysis framework

The report conducts fundamental analysis by combining second-quarter operating updates, owned and planned power capacity, project development timelines, AI lease negotiation progress, and partner capabilities, and uses a sum-of-the-parts valuation approach to separately assess the Bitcoin mining business, Bitcoin holdings, and the incremental power value available for AI deployment.

Methodology notes

• Valuation methodSum-of-the-parts valuation

  Separately values the mining business, Bitcoin holdings, and incremental power available for AI, then sums them.

  The Bitcoin mining business is valued at approximately 8x estimated 2027 adjusted EBITDA; Bitcoin holdings are valued at the current BTC price; incremental power available for AI deployment is valued at $3 million per megawatt, resulting in a $17 target price.

• Relative valuationEnterprise value multiple method

  Assesses the mining business using a multiple of enterprise value relative to estimated adjusted EBITDA.

  The report applies approximately 8x estimated 2027 adjusted EBITDA to MARA's Bitcoin mining business, reflecting the profitability and cyclical risks of the business.

• Catalyst analysisCommercial validation framework

  Uses actual AI contracts to validate the feasibility of converting power assets into AI infrastructure value.

  The first AI leasing or commercial contract would provide observable evidence on demand, pricing, development progress, and capital returns, and is therefore viewed as a potential re-rating catalyst.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• MARA.US

  The core subject of the report; the company's value is exposed simultaneously to Bitcoin mining, power assets, and AI infrastructure transformation.

  Strengths

  Owns and continues to expand large-scale power resources; management is advancing multiple AI commercial opportunities; Starwood provides mature engineering and development capabilities; the target price implies significant upside versus the current share price.

  Weaknesses

  The AI strategy has not yet been validated by signed commercial contracts; large projects have long development cycles and high capital requirements; traditional mining profitability is highly volatile.

  Comparison

  Compared with pure-play Bitcoin miners, MARA is attempting to increase the unit value of its power assets and diversify its business through AI and high-performance computing deployment.

  Risks

  Failure to sign AI contracts on schedule, delays in FERC approval, project construction execution falling short of expectations, external power partners reallocating capacity, environmental regulatory pressure, and equity dilution.

• BTC

  MARA's mining revenue, value of coin holdings, and sum-of-the-parts valuation are all closely tied to the BTC price.

  Strengths

  A rise in the BTC price can boost mining revenue, asset value, and liquidity.

  Weaknesses

  Price volatility amplifies instability in earnings and valuation.

  Comparison

  Even as the AI transformation progresses, BTC remains an important driver of MARA's near-term fundamentals and valuation.

  Risks

  A decline in the BTC price, an increase in network hashrate, and rising mining difficulty could compress profitability.

• AI and high-performance computing infrastructure

  The company plans to use new and existing power resources for AI capacity deployment, representing a potential second growth curve.

  Strengths

  Large planned capacity scale; Starwood has experience delivering more than 7GW of infrastructure; the Exaion platform can support sovereign-level AI and high-performance computing deployments in Europe and other regions.

  Weaknesses

  The business model, lease terms, customer quality, and capital returns still lack support from signed contracts.

  Comparison

  Compared with traditional mining, AI infrastructure may bring longer-duration and more visible contracted revenue, but construction standards and delivery requirements are also higher.

  Risks

  Customer signing delays, engineering cost overruns, grid interconnection and approval delays, mismatched technical specifications, and financing pressure.

Key data

• RatingMarket-PerformBernstein maintains the rating, applicable to a 12-month investment horizon.
• Target price17.00 USDImplies approximately 60% upside versus the $10.65 closing price on August 6, 2026.
• Planned power portfolioApproximately 4.2GWPotential scale achievable after the completion of recent transactions, excluding approximately 0.6GW of hosted capacity.
• Matagorda County potential capacity2GWAcquisition price of $600 million, linked to development milestones; an initial 1GW is expected before the third quarter of 2027, reaching 2GW in 2028.
• Long Ridge potential capacityApproximately 1GWThe transaction is still awaiting FERC approval and includes 505MW of generation capacity and approximately 200MW of capacity at the existing Hannibal site.
• Existing owned power assets1.2GWDistributed across markets such as Texas and Ohio, with approximately 0.7GW of sites already in operation.
• Hosted capacityApproximately 0.6GWThis capacity is not included in the planned owned power portfolio shown in the report.
• AI agreement targetAt least two by the end of 2026Management reiterated the target and stated that commercial negotiations at multiple sites remain ongoing.
• 2025 actual revenue907 million USDData from the report's financial forecast table.
• 2026 estimated revenue988 million USDData from the report's financial forecast table.
• 2027 estimated revenue1,370 million USDData from the report's financial forecast table.
• 2027 estimated EBITDA544.24 million USDHigher than the 2026 estimate of $270.27 million and the 2025 actual figure of $132.06 million.

Impact & implications

If MARA signs economically attractive AI contracts before year-end, its power portfolio could be re-priced from traditional mining resources into scarce AI infrastructure assets, improving revenue visibility and valuation multiples. Conversely, contract delays, project approval obstacles, or rising development costs would weaken the transformation narrative, leaving the company's valuation still dominated by Bitcoin prices, network hashrate competition, and the mining cycle.

Risks

• AI commercial agreements may not be signed within the timeframe expected by management, delaying the re-rating catalyst.
• The Matagorda County and Long Ridge projects may face approval, development schedule, construction cost, or grid interconnection risks.
• The company relies on external power partners, and if partners redirect capacity to other projects, output may be affected.
• Continued intensification of network hashrate competition while the company's capacity expansion is insufficient could dilute its mining share.
• Volatility in Bitcoin prices, mining difficulty, and energy costs could significantly affect earnings and valuation.
• Energy consumption and the use of thermal power may trigger pressure from environmental groups, regulators, and communities.
• Large-scale investment in power and AI infrastructure may create financing needs and equity dilution.

What to watch

• Whether at least two AI leasing or commercial agreements are signed before the end of 2026.
• The customer, capacity, lease term, pricing, capital investment, and return terms of the first AI contract.
• Batch Zero approval for the Matagorda County project and progress toward the initial 1GW target in the third quarter of 2027.
• The FERC approval result for the Long Ridge acquisition and the timing of transaction completion.
• The pace at which Starwood partnership projects move from lease negotiations to construction and delivery.
• Order progress for the Exaion platform in markets outside Europe.
• Owned power asset operating capacity, Bitcoin mining capacity, and the realization of EBITDA from 2026 to 2027.