Summary MARA Holdings remains primarily a Bitcoin miner and treasury, with its valuation closely tied to the price of Bitcoin rather than AI infrastructure growth. While management has announced partnerships and acquisitions related to AI and high-performance computing, these initiatives are still early and unlikely to generate meaningful revenue in the near term. Mining economics have become more challenging as energy costs per Bitcoin have risen significantly, putting pressure on margins when Bitcoin prices decline. Given limited near-term catalysts outside of Bitcoin price appreciation, MARA's risk/reward profile appears balanced, leading to a Hold rating. Investment Thesis While several Bitcoin miners have pushed aggressively into AI and HPC infrastructure, MARA Holdings, Inc. ( MARA ) has been slower to move. After a summer of huge deal announcements from other Bitcoin miners such as IREN Limited ( IREN ) and Cipher Digital Inc. ( CIFR ), it shows MARA is late to the shift. The company is still working in that direction, but for now, MARA remains a Bitcoin treasury first with limited near-term AI/HPC upside. A transition announcement was made in Q4 of last year, but they haven't made much progress. The first step was made in August when they signed an agreement to take a majority stake in Exaion, a developer and data center operator. The company has also announced larger deals, such as a joint venture with Starwood Capital. However, this is unlikely to be a contributor to near-term revenue. Although mining operations have become more efficient , Bitcoin mining has changed for everyone after the April 2024 halving event. This has caused energy costs per Bitcoin mined to rise significantly across the industry, with the block reward going from 6.25 BTC to 3.125 BTC. In 2024, the energy cost for MARA to mine Bitcoin was $28,201, which grew to $48,611 by Q4 2025. This roughly 72% increase can largely be attributed to the post-halving environment, where miners must generate twice the compute to earn the same amount of Bitcoin as before the halving event. Year Bitcoin Holdings 2021 8,115 2022 12,232 2023 15,126 2024 44,893 2025 53,822 Besides mining, MARA has also adopted a large Bitcoin treasury, which grew with large purchases in 2024 and 2025. In my opinion, this company is not much more than a Bitcoin treasury. The risk/reward doesn't appear to be very attractive, and unless you want leveraged Bitcoin exposure, I'd look elsewhere. That's why I'm giving this stock a Hold. Too Late To Transition? MARA has been behind when it comes to an AI/HPC transition. Competitors like IREN Limited ((IREN)) and Cipher Digital Inc. ((CIFR)) spent 2024 getting contracts with hyperscalers and started building out capacity. Meanwhile, MARA just began forming the idea. These miners made deals with Amazon, Google, and Microsoft throughout 2025, while MARA deployed its first 10 inference racks in Q3 of the same year. Enterprise GPUs, such as Nvidia's H100, are often rented for between $3 and $7 per hour. This implies revenue of $26,280-$61,320 per GPU per year if running 24/7. For MARA, that means revenue of roughly $2.1 million to $4.9 million annually, assuming eight GPUs per rack. This suggests that these 10 racks are a pilot program and not yet a significant revenue driver. The Exaion acquisition was the first move of the transition, but it is quite a small step toward hyperscaler contracts. After closing the deal for $168 million, they now have access to about 1,250 GPUs across four data centers in Europe, which is small compared to most. The real upside for them is using Exaion to build out new centers and transforming current mining operations, something that could take years. A collaboration with MPLX was announced in late 2025, allowing MARA to build with an initial capacity of 400 MW and potentially scale up to 1.5 GW. This allows them to use MPLX's cheap electricity, but it was a letter of intent with no timeline mentioned. This could also face constraints beyond securing low-cost energy. Large-scale data center developments such as this one require interconnection approvals, transmission upgrades, and permits before power can be delivered to new facilities. These interconnection queues have grown as demand from hyperscalers has increased. This means that even if the MPLX partnership goes beyond a letter of intent, bringing the full facility online would likely take years, depending on grid availability and infrastructure upgrades. Another partnership with Starwood Capital was announced, where selected mining sites would turn into AI data centers. This is a stronger partnership, since Starwood manages over $125 billion in assets and brings data center expertise through Starwood Digital Ventures. It's important to note this is a JV, where upside is split between the companies. Although these deals do show proof of MARA's intentions to move into HPC/AI, revenue from these deals is likely years out. While MARA is behind its peers in transitioning, there could be some advantages to a later entrance. Early adopters deployed large clusters during the peak of the buildout, when GPU, notably Nvidia Blackwell, supply was constrained and prices were elevated. Companies now have the opportunity to deploy that same next-gen hardware, but at more normalized market prices now that things have cooled off. If MARA's buildout moves with the shift, it could lead to lower initial hardware prices relative to earlier adopters. Bitcoin Dependency Without the AI/HPC possibility, MARA is simply a Bitcoin miner and treasury. As of now, its financial performance is tied closely to the price of Bitcoin. In their most recent earnings report, revenue came in at $202 million, below the consensus of $252 million and down ~6% YoY. This decline came mostly from Bitcoin's drop, but rising operating costs, such as energy price increases YoY, also contributed . Revenue did grow 38% YoY to $907 million, helped by all-time highs in crypto. The difference between the 38% annual revenue growth and the ~6% decrease YoY can be explained by MARA's hash rate increase and Bitcoin's price movements. During Q4, the company's energized hash rate increased from 60.4 EH/s in Q3 to 66.4 EH/s in Q4. Normally, this growth would contribute to higher output and revenue. However, Bitcoin prices fell throughout the quarter, causing revenue generated per hash rate to decline. Fleet efficiency remains competitive, with MARA operating at around 18.6 J/TH. However, improvements in scale have not been able to offset margin compression. The purchased energy cost per Bitcoin mined rose from $39,235 in Q3 to $48,611 in Q4 . This highlights the dependency of Bitcoin's price on mining, where efficiency improvements are unable to counter the price. MARA is still one of the larger miners in the sector. For comparison, a close competitor such as CleanSpark operates at 50.0 EH/s of hash rate. While MARA's scale does provide some advantages, many peers have expanded mining capacity while securing AI/HPC hosting contracts. Debt Increase and Dilution Throughout the volatility of the last year, MARA's debt has grown. The company now has $3.6 billion in debt, making the Bitcoin-to-debt ratio ~1.04x at a Bitcoin price of $70K (53,822 coins), meaning they now owe about as much as they hold in Bitcoin. The Bitcoin price at which MARA's holdings equal debt is $66,900 per coin. While the long maturities and low interest of their debt reduce near-term pressure, this shows how closely MARA's financials are tied to the Bitcoin price. The major parts of debt include a $1 billion offering of 0% convertible senior notes due 2030 as well as an upsized $950 million offering of 0% convertible notes due 2032. While the structure of the debt is nice, it does introduce dilution risk if the company's stock price appreciates enough. The 2032 notes carry a conversion price of $20.26, and the 2030 notes carry a conversion price of $25.91. If progress is made on the AI/HPC transition or Bitcoin returns to all-time highs during the conversion period, the price would likely be high enough for conversion. Dilution has been persistent throughout the last year, with the diluted share count increasing by 14% YoY, which was an improvement from 2024, when diluted shares outstanding grew by about 62%. The convertible notes, if prices reach the levels needed for conversion, are not included in today's share count and could cause future dilution. These are some basic elements of the DAT flywheel, where equity and/or debt are issued, and the company hopes that Bitcoin rises fast enough to offset the cost. Valuation In my opinion, a fair valuation for MARA is to value it like a DAT. This would be by using a market cap-to-net-asset-value (mNAV) multiple, a traditional metric for DATs, calculated by dividing market cap by net asset value. This represents the ratio of the two and allows investors to see if the company is trading at a premium or discount to its assets. When Bitcoin was at an all-time high, many DATs traded at a premium to their net assets. However, as prices have fallen, most DATs have settled around an mNAV of 1x. This implies the market is valuing these companies as passive Bitcoin holding vehicles. While they do get some revenue from mining, those margins are now very thin, and the majority of their assets are in Bitcoin. MARA is trading near a 1x mNAV multiple, showing investors lack confidence in the HPC/AI transition and view it as a DAT. If the mNAV were below 1x, it could present an entry point as investors gain exposure at a discount. On the other hand, some catalysts for a sustained mNAV above 1x would be a significant rally in Bitcoin prices, hyperscaler hosting agreements, or clearer progress made on large-scale data center development. Analyst sentiment has been trending bearish, with Cantor Fitzgerald cutting to $11 from $21, Clear Street from $16 to $9, and Jefferies from $13 to $9. While the median target is still high at $16.36, many revisions are yet to be released after Q4 results. In my view, the stock's current valuation is fair. With little HPC/AI upside coming in the near term, thin-margin mining operations, and a lower Bitcoin price, MARA doesn't have much to offer unless you're betting on a rise in Bitcoin. My fair valuation of MARA is an mNAV of 1x, where the stock moves with the value of its NAV, which is primarily in Bitcoin. That's why I'm rating this stock a Hold, since it's already trading in that range. Risks to the Thesis The biggest risk is unexpected Bitcoin upside. If Bitcoin rallies back to ATHs, treasury mNAVs could potentially expand back to premiums. This could also help mining operations, where margins are currently thin. If timelines for data center expansion advance or new deals get made, the market could be convinced MARA deserves a premium for HPC/AI upside. Conclusion MARA has consistently moved with industry hype. When mining became popular, they went into it. That strategy then transitioned into being a Bitcoin treasury and now HPC/AI. While they may be the right moves, the company has certainly been late to shift. At its current mNAV around 1x, the market isn't pricing any HPC/AI upside, which I believe to be the correct valuation. In my opinion, more progress and announcements have to be made before they are deserving of a higher multiple. If you're looking for a former miner turned HPC/AI, I would look elsewhere. MARA has the appearance of a leveraged Bitcoin vehicle, and unless you're looking for that, this isn't a stock I'd recommend. This is why I'm rating it a Hold, with a target multiple of 1x mNAV.
