Eyes On Q3: The AI Catalyst That Could Wake Up Markets On MARA Holdings

Summary MARA Holdings has lagged peers YTD, while IREN quintupled on its AI pivot. I see it as an asymmetric bet, wrongly seen by the market as a Bitcoin treasury ￰0￱ a 1.20x mNAV premium versus MicroStrategy's 1.34x, MARA's ~53,000 BTC holdings worth $5.9B justify a higher multiple given its mining inflows and AI potential. MARA's mining business alone merits a ~$5B valuation, projecting 64,000 BTC mined through 2048 worth ~$7.4B at current prices or up to ~$48B+ in a bullish Bitcoin case. A two-phased AI pivot, contracting 200 MW power capacity then leveraging Exaion's partnership for GPUs, could add $405 Mln Yearly in gross margin and represent a catalyst for the stock.

I rate MARA a Strong Buy with 2x from mining and 3x from AI upside, though dilution, AI bubble risks, and BTC dependence remain significant risks. I last covered MARA Holdings, Inc ( MARA ) in July , arguing the market was sleeping on the true potential of this Bitcoin ￰1￱ stock is roughly flat since my last coverage in July, underperforming the broader ￰2￱ holders (including myself) may be especially frustrated by the fact that many of MARA’s peers have significantly outperformed it, as the chart below ￰3￱ vs. peers, YTD (Seeking Alpha) The most prominent example of outperformance is that of IREN Limited ( IREN ). This crypto miner began a strategic pivot to AI cloud services in July 2025, leveraging its existing GPU-powered data centers for high-performance ￰4￱ reporting its first AI Cloud revenue for FY2025, the stock more than quintupled against its levels in ￰5￱ miners that did not explicitly pivot to AI have significantly outperformed MARA.

CleanSpark, Inc. ( CLSK ), for example, is outperforming MARA YTD by a factor of more than ￰6￱ company achieved so by avoiding shareholder dilution and by contracting 1 GW power capacity, which helped with ￰7￱ this article, I will explain why I believe MARA may be on the cusp of a strategic pivot that could ignite new growth for the company and boost its (so far, lagging YTD) stock performance. MARA: trading like a Bitcoin Treasury Company Readers may have noticed how in my previous chart I included Strategy Inc. ( MSTR ) for comparison with MARA and other miners.

Strategy, a company I recently covered , is a Bitcoin proxy investment that is underperforming MARA itself, with flat returns ￰8￱ is not by chance. I think the market is seeing both MARA Holdings and Strategy as Bitcoin treasury ￰9￱ if they are not alike. Personally, I define a Bitcoin treasury company as a company whose main purpose is to acquire and hold Bitcoin long term, believing in the future of this cryptocurrency as a global reserve ￰10￱ that sense, a Bitcoin treasury company’s other operations become as a secondary strategic ￰11￱ officially announced it would start a BTC “hold” strategy back in January 2021 (previous to that, they resorted to sell most of the Bitcoin they mined to finance themselves).

The company has also been very vocal about how important holding Bitcoin long term is for the company, celebrating reaching 50,000 in their treasury in ￰12￱ my view, this amounts to a continued Bitcoin treasury strategy. I think this strategy has so far punished MARA for what concerns market ￰13￱ currently holds ~53,000 Bitcoins, which are worth ~$6 Bln at the time of writing (with Bitcoin at ~$115,000 per coin). With a market capitalization of ~$7.2 Bln (Seeking Alpha data), the company trades at an mNAV premium of ￰14￱ is less than Strategy’s mNAV premium , which stands at ~1.34X at the time of writing. I find this difference very difficult to justify and worthy of calling MARA an asymmetric bet in the stock ￰15￱ of all because MARA, contrary to Strategy, is one of the largest Bitcoin miners on the ￰16￱ means there is a relatively certain future inflow of Bitcoins that MARA will generate and must be accounted for by investors.

Strategy, on the other hand, can only resort to issuing debt or diluting shareholders to monetize their mNAV premium and acquire more ￰17￱ other reason why I find MARA’s valuation very cheap has to do with the potential of the business in case of an AI (and energy) pivot, something I believe MARA, contrary to Strategy, could do and that I will cover ￰18￱ point is that MARA’s focus on a BTC treasury strategy has punished the stock and blindsided the market to view this company purely as a box of Bitcoin trading at a slight ￰19￱ cannot be further from the truth, as I will now outline in my next sections. MARA’s BTC mining business is worth ~$5 Billion by itself As one of the largest Bitcoin miners in the world, I think much of MARA’s value lies in how much Bitcoin the company can be expected to mine in the next few years.

I have touched on this topic in my previous coverage for this stock, and previously calculated the following Bitcoin metrics: I expect MARA to produce ~27,000 Bitcoins until the 2028 halving (20 BTC per day, multiplied by 1349 days until mid-April 2028). I then expect the company to produce ~16,000, ~9700, ~ 5700, ~3500, and ~2100 Bitcoins for the following five halvings (ending in 2048). The total is 64,000 ￰20￱ current market prices NDR: Prices from March 2025, 64,000 Bitcoins are worth $5.7 ￰21￱ my Bitcoin valuation model has an expected value between $112,500 and $200,000 per coin, I would assign the value of these Bitcoins at between $7.2 Billion and $12.8 ￰22￱ Bitcoin mature into a global reserve asset, matching gold's market capitalization at roughly $750,000 per coin, these Bitcoins would be worth $48 Billion.

NDR: Assumptions and quotes from my March 2025 article At current Bitcoin market prices, the 64,000 Bitcoins that MARA will mine in the upcoming two decades are worth ~$ 7.4 ￰23￱ an investment in MARA inevitably makes sense only assuming a bull case for Bitcoin (more on this inherent risk later in this article). Assuming a (conservative, for a bullish case ) long term price per Bitcoin of ~$ 250,000 per coin, and a 30% long term net profit margin for MARA, this stack of Bitcoin is worth ~$ 4.9 Billion in net profit, which represents roughly MARA’s market capitalization at the time of ￰24￱ other terms, I think MARA's BTC mining business is worth a 2X from share price levels at the time of ￰25￱ note, these assumptions are speculative and can only serve as a rough estimate of how much MARA's mining business is ￰26￱ example, MARA is unlikely to ever monetize its mining business by simply liquidating acquired Bitcoin and "pocketing" a 30% net margin as I assumed in my ￰27￱ would be inconsistent with their BTC "hold" ￰28￱ are, in my view, more likely to maintain a long term BTC hold strategy and rather focus on monetizing an AI pivot.

What’s important for the sake of my thesis, however, is that today’s ~ $1.2 Billion market valuation of MARA’s entire mining business (the company market cap minus the market value its BTC holding) is, in my view, clearly ￰29￱ can make their own conclusion about how much MARA’s future inflow of ~64,000 Bitcoin is worth. What’s reasonable, I believe, is that it should be worth significantly more than $ 1.2 ￰30￱ could pivot to AI with a two-phased approach To understand what MARA’s AI pivot could look like, we can benefit from the fact that many of its direct peers have already done such a ￰31￱ is why, for a start, it’s important to understand the difference between MARA and companies like IREN and ￰32￱ that purpose, I have compiled a simple table with key figures and ￰33￱ for the below data points are MARA’s BTC report and Q2 shareholders letter , IREN’s FY25 report and Yahoo Finance , unless specified ￰34￱ Holdings IREN CleanSpark Power Capacity 1.1 GW owned; 760 MW active for mining 810 MW online; 2,910 MW secured total 1.03 GW under contract; 808 MW utilized Cost per KWh $0.04/kWh $0.035/kWh $0.056/kWh Data Centers geographical footprint US-focused with a presence in Europe via Exaion partnership Canada (Prince George: 50 MW air-cooled) & Texas (100k+ GPUs), optimized for dense AI Exclusively US portfolio (e.

g., Tennessee via GRIID acquisition, College Park GA for AI) AI/HPC Hardware Early stage, no major owned GPUs yet 23,000 GPUs already deployed Nascent, no GPUs deployed ￰35￱ the above in mind, I believe MARA could follow a 2 steps approach to pivot away from sole crypto mining: Start contracting power capacity for AI applications, meaning committing access to a defined volume of electrical power through long-term agreements with utilities or energy ￰36￱ is a similar approach to what CleanSpark has recently ￰37￱ the revenue from contracting power capacity to acquire GPUs and rent them out directly for AI ￰38￱ is the same approach IREN ￰39￱ reason behind this two-phased approach is that, as per my table above, MARA does not own or deploy a significant amount of ￰40￱ company rather uses so-called “ASICs” (Application-Specific Integrated Circuits), which are custom built chips designed specifically for mining ￰41￱ ultimate goal of this strategy would be to rent out directly GPUs for AI cloud services or rentals of high-performance ￰42￱ is what IREN has been doing, thanks to its 23,000+ GPUs in active ￰43￱ Exaion partnership: a first step towards AI, unnoticed by the market?

A signal that MARA may be actively thinking about an AI pivot is its recently expanding footprint in ￰44￱ recently acquired a majority stake in Exaion, with the option of buying the entire company in the ￰45￱ is a European based company that provides secure cloud and AI infrastructure, formerly a subsidiary of EDF, a large French energy ￰46￱ operates 1,250 GPUs across its four data centers at the time of ￰47￱ is unlikely, in my opinion, that these GPUs could be put to any other use in the immediate term (they are already utilized in Exaion’s business, which will continue to serve existing clients, including EDF). However, I find it interesting that management used the following wording related to the acquisition: Exaion would transition to a larger-scale international commercial ￰48￱ integrating Exaion’s platform, operations, and engineering expertise, MARA expects to expand its capabilities into AI/HPC infrastructure development In my view, this is a signal that management is thinking about an AI ￰49￱ the context of my two-phased approach, this is equivalent to preparing for step 2 (the most profitable one).

With Exaion, MARA has access to the HPC infrastructure needed to deploy ￰50￱ MARA to complete its full acquisition of Exaion, it could accelerate its move to step 2 (i. e. renting out GPU computational power) by leveraging Exaion’s existing ￰51￱ the markets have not reacted to news of the Exaion partnership so far, I think this may act as a catalyst if MARA’s management outlines their plan in the upcoming ￰52￱ AI pivot could represent a further ~3X upside for MARA Assigning an exact monetary value to an AI pivot for MARA is a difficult exercise, considering the company has not yet announced any plans in that ￰53￱ a thought starter however, I will still outline what could be the financial impact of MARA starting to contract out power capacity for its first year of ￰54￱ assumptions are as below: 200 MW of dedicated AI capacity to rent out (what’s left of 1.1GW of total capacity, less 760 MW currently utilized for mining) A rate of $220 / Kw / Month, which is the midpoint of 2025 global data center colocation pricing .

A utilization rate of 90%, accounting for setup and downtime for the first year, resulting in 8,760 hours / year of operations A Power cost of $0.04/KhW, MARA’s own blended rate as of Q2 2025 For simplicity, I will exclude other OPEX cost With the above assumptions in mind, I estimate a potential additional Gross Margin of $ ~400 Million, as per calculations ￰55￱ Revenue 200 MW × 1,000 kW/MW × $220/kW/mo × 12 mo $528 million Adjusted Revenue (90% Util.) 90% of $ 528 Mln $475 million Power Costs 200 MW × 1,000 kW/MW × 8,760 hrs × $ 0.04/kWh $70 million Net Earnings (Gross Margin) $ 475 Mln- $ 70 Mln $405 million Exactly what value the market may assign to such a new business (if it is ever announced) is a difficult and somewhat speculative exercise.

First, because the market is likely to look well beyond the additional gross marginality of a single year of ￰56￱ importantly, because MARA’s own financials are the results of mark-to-market accounting obligations in relation to the BTC they ￰57￱ makes isolating and benchmarking this new revenue stream a difficult exercise at best. A more interesting exercise can be, once again, looking at the cases of IREN and CleanSpark. IREN, CLSK, P6M (Seeking Alpha) CleanSpark roughly doubled year-to-date (+115%), while IREN performed significantly better with a more than 800% return YTD, as per chart above. I think a bull run for MARA could manifest as something in between these two ￰58￱ face value, MARA’s AI pivot is more comparable to CleanSpark: simply contracting 200 MW of extra power capacity for AI use cases.

However, I think Exaion’s partial acquisition may drive further gains in the market on expectations of an accelerated roll out to phase 2 using its ￰59￱ summary, I would expect an AI pivot for MARA to represent a ~3X upside for the stock, purely based on the price action of MARA’s direct peers and my two -phased AI pivot theory. Obviously, such a bull scenario could only manifest as long as the market remains convinced demand for AI applications (and data centers) is going to continue and/or ￰60￱ is something I will cover in my next section, together with other risks linked to investing in MARA. Risks: shares dilution, AI bubble and crypto winter A bet on MARA remains a leveraged bet on Bitcoin and, if a pivot were to come, on ￰61￱ market is likely to assign a premium valuation to MARA only as long as AI remains a hot investment theme, and Bitcoin continues its secular bull ￰62￱ this sense, investors should carefully consider whether they deem likely that the current climate of euphoria on both topics is going to ￰63￱ who are bearish on Bitcoin or believe that AI is a bubble should of course stay out of an investment in MARA Holdings altogether.

A further potential risk of an investment in MARA is linked to shareholders ￰64￱ MARA’s management has been heavily diluting shares, as the chart below ￰65￱ forward, I expect this trend to continue, considering that the company’s “HOLD” strategy means MARA needs to incur in dilution to finance its ￰66￱ share count (MacroTrends) This represents a significant headwind for MARA, which can only be compensated by strong retail demand for the ￰67￱ I see MARA as an asymmetric bet in the market, it is a highly speculative and high risk bet, depending both on execution (management announcing an AI pivot) and overall market ￰68￱ interested in entering a position in MARA should bear this in mind.

Conclusion: MARA is an asymmetric bet on AI and BTC mining MARA has lagged behind Bitcoin and its crypto mining peers year-to-date. I think this may be about to change, as a result of two elements: A two-phased pivot to AI, first focusing on renting out power capacity and then acquiring and renting out GPUs, leveraging the recent deal with Exaion. A continued Bitcoin bull run that may awaken the market to MARA’s mining business long term potential, which I estimate will mine ~64,000 Bitcoins in the next two ￰69￱ two elements respectively represent a 2X and 3X upside from current prices, according to my ￰70￱ potential for asymmetric returns brings me once again to recommend a “ STRONG BUY ” for MARA Holdings.

I will look at signs of an AI pivot in the next earnings release, which is set for early November.