Summary MARA Holdings (MARA) remains highly exposed to Bitcoin price movements, with operating results still largely driven by the underlying commodity rather than stable, company-specific cash flows. The ongoing transition toward vertically integrated mining should improve unit economics over time, but at current production levels, profitability remains sensitive to Bitcoin prices. Management’s push into AI and high-performance computing adds potential long-term optionality, though the initiative is still early and not yet reflected in reported revenue or cash flow. With cost reductions, production stability, and diversification still in progress, the stock appears more fairly valued than obviously mispriced at present levels. MARA Holdings ( MARA ) is down around 50% over the last year as it's been ramping up CapEx at exactly the wrong time: when bitcoin is down 30% over the last quarter. I want to take a look at MARA to see if there's value here at this lower price. The issue with MARA is that it's not that easy to analyse by just looking at financials. Things like mark-to-market bitcoin price moves have an impact on the P&L, along with the current expense and CapEx ramp-up. So my goal with this analysis is to normalise these impacts and do a sum of the parts analysis to derive the current enterprise value of MARA and see if it can be justified. Current Bitcoin The easiest way to start a sum of the parts is by looking at the value of the current bitcoin on the balance sheet and comparing that to the enterprise value. Bitcoin is $87k at the time of writing. As seen in the Q3 investor presentation , MARA currently holds 52,850 bitcoin. Q3 2025 Investor Presentation 52,850 * $87k = $4,597,950,000 The current market cap is $3.76 billion, so we can add back financial obligation, minority interest, and subtract cash to get enterprise value. Financial obligations - add back: Current portion of LT debt: $350 million Current portion of lease obligations: $1.5 million Long-term debt: $3,247.6 million Capital Leases: $43.4 million Total: $3,642.5 million Minority Interest - add back: $16.7 million Cash & Equivalents - subtract: $826.4 million Total add back = $2832.8 million $3.76 billion + $2.8328 billion = $6.5928 billion = Enterprise Value We can subtract the current bitcoin holdings to get what's left over, which is $1,994,850,000, so let's call it $2 billion. The primary goal now is to see if the rest of the business can be valued at over $2 billion; if so, then MARA is undervalued and vice versa if not. Revenue & Expenses of Production The key business segment, besides just holding bitcoin akin to a bitcoin treasury, is the production side of the business. Q3 2025 Investor Presentation Looking at the Q3 investor presentation, 2,144 bitcoin were produced. 2,144* $87k = $186,528,000 in quarterly revenue expected at the current bitcoin price. The expense side is where things start to get tricky. This is because MARA is transitioning from an external hosting system to an internal one. Previously, MARA would pay third-party vendors to host, with the vendors covering things like energy, labour, maintenance, site operations, and local management. This was an asset-light model where MARA outsourced these needs. Now MARA is transitioning to a vertically integrated model where it owns and operates the sites itself. This means it must now hire additional staff, run facilities, maintain miners, manage power infrastructure, run data centres, handle compliance, safety, plus logistics, and so on. The current part of the transition that MARA is in, opex appears high because it's still paying many of the third-party fees whilst simultaneously paying for the new internal team as it ramps up. So you get a duplicate cost structure where margins look worse than they are when normalised. This can be seen when looking at the P&L comparing Q3 2024 to Q3 2025 on the Q3 2025 10Q : MARA Q3 2025 10Q As seen above, opex has gone up over the last year. So the key now is to figure out what this all normalises to once MARA transitions to a fully vertically integrated model, where it doesn't have to pay the same level of third-party hosting fees. The four key expenses to look at: Purchased energy cost Operating & maintenance cost Third-party hosting & other energy costs General & administrative cost The company has highlighted in its Q3 2025 shareholder letter that the overall cost per petahash is down 15% year to date: MARA Q3 2025 shareholder letter Translating this to cost per kWh is $0.04, and for the cost per bitcoin is $39,235. Now it's difficult to quantify exactly where purchased energy cost will be in the future, but we can make a pretty good estimate based on the information the company has given us and at least aim to be directionally correct, even if not spot on. Third-party hosting and other energy costs consist of co-location services related to third-party hosted sites and energy expenses related to mining other digital assets. Third-party hosting and other energy costs in the quarter were $75.7 million compared to $63.7 million in the prior year period, an increase of $12.0 million. The increase was primarily due to the addition of energized miners and the expansion of third-party hosted facilities compared to the prior year period. Our pivot from asset-light to vertically integrated helped to reduce our electricity cost per coin to one of the lowest in this sector. As we transition towards a more owned and operated model, phase out third-party hosted contracts over time, and bring low-cost sites like wind farms online, we expect costs on a unit basis to continue to improve. In essence, there's ambiguity here since they didn't provide exact forward guidance on where they expect the purchased energy cost to be, but they did confirm the trend will continue over the next couple of quarters. Just based on the current reduction trend continuing, I think we can expect around a 20% reduction in this line item once these costs are further reduced and once there's less reliance on third-party hosted sites. Last quarter's purchased energy cost line item was $43.080 million, so a 20% reduction is $34.464 million. For operating and maintenance costs, it includes: on-site labour, miner maintenance and repairs, cooling systems, electrical infrastructure upkeep, and site operations & security. The key thing with this expense to remember is that O&M rises when you internalise operations, but it will fall once bitcoin utilisation and scale normalise. The other thing is that it's a lot more akin to a fixed cost, so it has operating leverage as it scales. I'm going to keep O&M costs the same, because it's largely fixed, and the real benefit of it is only seen if bitcoin production goes up. It's a similar story with general and administrative costs. G&A stays the same regardless of how much you produce, so it just carries over as a fixed cost. Quarterly expense for both: O&M: $26.31 million G&A: $85.296 million This leaves us with third-party hosting costs, which is where the largest expense cut is expected. Two years back, in December 2023, Nasdaq put out a press release saying: Currently, Marathon’s Bitcoin mining portfolio consists of 584 megawatts of capacity, 3% of which resides in sites that are owned and/or operated by the Company, and 97% of which is hosted by third parties. Following the close of this transaction, Marathon’s Bitcoin mining portfolio will consist of approximately 910 megawatts of capacity, 45% of which will reside on sites directly owned by the Company, and 55% of which will be hosted by third parties. So there was already clear directional data on where things were headed. I think it's safe to make a presumption that a reduction will continue and that we could see a 40% reduction in this line item. I say this since we already know that per petahash per day improved 15%. And we know that O&M and G&A are largely fixed costs, so the bulk of improvement has to come from the third-party hosting cost. Current quarterly third-party hosting costs are $75.664 million, so a 40% reduction is $45.3984 million. Adding this all up makes for a quarterly operating expense of $191.4684 million, plus we still need to add other miscellaneous expenses. These are taxes other than on income, early termination expenses, and research and development. These are all around 16 million. So let's say the total quarterly operating expense is around $207 million. With quarterly production revenue at $186,528,000, they are underwater; although there are some additional revenues such as interest income from loaning out bitcoin, which last quarter was $17,689. In all, once accounting for all this, at the current production levels and $87k price per bitcoin, it's a breakeven EBITDA under our expense assumption. So you're really paying for the possibility that they grow production to take advantage of operating leverage, the bitcoin price goes up, or that they optimise expenses to the point that they can virtually get rid of all third-party hosting. It's possible they grow production or that they optimise third-party hosting cost, but I wouldn't want to bank on it. AI & HPC Call Option There's an AI call option that seems to be all the rage, but frankly, I just don't get it since the time to mass recurring revenue scale is far away and profitability even further. The idea behind it is that the energy and data centre infrastructure they have to mine Bitcoin can be repurposed to perform other types of compute work, including AI inference workloads. In pursuit of this vision, Marathon deployed its first ten AI inference rack systems at the Granbury, Texas, facility in 2025 and announced an acquisition of an initial ~64% equity stake in Paris‑based Exaion, a provider of tier 3/4 data centre and compute services. I view this as pure optionality for a few key reasons: Lack of Monetised Compute Revenue Uncertain Utilisation, Pricing, and Margins Competitive and Capital Intensity Timing Risk Until we see these issues addressed, it's just a form of optionality, but it won't get to the point where it'll scale with recurring revenue and profitability. Dilution Something else that's been an issue is how reliant MARA is on diluting equity holders to raise capital: Seeking Alpha Source It's pretty much been straight up and to the right. Dilution works when your stock is richly priced relative to what you're investing in, and there's a clear and immediate payout. That isn't the case with MARA. The stock is down heavily, which means that to get the same level of capital, they need to dilute even more stock. Plus, mining operations are breakeven at $87k bitcoin, per earlier calculations. And AI & HPC is just optionality that may or may not pay off with long-term sustainable profitability. So there is no clear payoff in the immediate term, yet shareholders are getting heavily diluted. Just over the last year, outstanding shares have been up around 27%. Now, with a lower stock price and a higher cash burn rate, they would need to dilute even more stock. Takeaway To take this full circle, what I posited originally is that the enterprise value is $6.5928 billion, so there's almost $2 billion of value that is blue sky above their bitcoin holdings, and we'd need to see if the rest of the business justifies that. The major issue I see is that everything else is just a call option right now. The bitcoin production side of the business is at breakeven with current bitcoin prices, even considering the operational improvements that will be made. Similar story with the AI side of the business in that it's just optionality. The $2 billion gap could be justified if you believe: Bitcoin production will go up Large-scale operational improvements will reduce operating expenses further Bitcoin prices will go up The AI business will generate significant recurring revenue soon But I simply don't see these right now, and I don't want to invest in something that's just a bitcoin proxy, which is what MARA is trading like, because so many of the economics are tied back to the price. If it were a lower-cost producer with a clear path to scale production, then that would be a different story, but it isn't right now. For these reasons, I've put a hold rating (neutral outlook) on MARA.
