Summary Marathon Digital (MARA) remains highly leveraged to bitcoin's price, lacking long-term HPC leasing deals that have stabilized peers’ revenues. MARA’s acquisition of Exaion, as well as their partnership with Starwood, signals a shift toward building its own HPC cloud offerings, aiming for higher upside but increased execution risk. Access to low-cost energy is a strategic focus for MARA, potentially supporting margins as it competes with aggressive HPC players like CoreWeave and Nebius. MARA is positioned as a high-risk, option-like play: substantial upside if bitcoin prices and HPC ambitions succeed, but significant downside if both falter. There have been a lot of bitcoin miners in the last 3-6 months that have been making deals with other companies (several with Fluidstack) where the bitcoin miner leases out there infrastructure and energy access for many years. In essence, companies are paying these bitcoin miners, so they can have instant access to a large amount of hardware and infrastructure to give a cloud offering for High-Performance Computing (HPC). If Fluidstack and others were to build out this amount of capacity on their own, it would take them a very long time, and the need for HPC for many of the hyper-scalers is right now, not 2-5 years from now. For the bitcoin miners it also makes sense, as they have been sitting on a lot of computing infrastructure that was built up as the price of bitcoin ( BTC-USD ) was going up. The economics have turned to where the compute cycles are starting to be worth more to support AI and other HPC needs than they are to just mine bitcoin. It also has provided a lot of the bitcoin miners with a lot more vision into their future revenues, with less overall risk. Here are a few of the deals that have been signed: In October 2025, TeraWulf ( WULF ) made a deal with Fluidstack In December 2025, Hut 8 ( HUT ) signed a potential $7B lease agreement with Fluidstack to provide 245 MW of IT capacity Core Scientific ( CORZ ) made a deal with CoreWeave ( CRWV ) In the Fall of 2025, Cipher Mining ( CIFR ) made both deals with Amazon ( AMZN ) and Fluidstack The Long Fall for MARA Before the latest earnings call on the evening of February 26th, MARA Holdings ( MARA ) was not one of the companies that has signed such a deal to lease out their hardware to other companies. This had led MARA to feel the full weight of the drop of bitcoin, just as their peers that did make deals turned out to be very well timed for the most part. This is just how bad it has been for MARA in comparison to some of these peers over the last year: Data by YCharts You'll see that as bitcoin was going up for a lot of 2025, so too were the prices of all the bitcoin miners. However, when the price of bitcoin started to fall in October, most of the miners that had made deals promising out their hardware held mostly steady, whereas MARA fell over 67% peak to peak. Now, this isn't likely entirely because MARA didn't make a deal similar to these other companies. MARA was also much more coupled to the price of bitcoin due to the high amount of bitcoin they carry on their balance sheet. Where many miners sell majority of the bitcoin they mine, MARA has decided to go the other route. They have not only kept what they mine, but have leveraged up and bought additional bitcoin, such as when they announced in July of 2024 that they had bought $100M worth of bitcoin. Luckily, most of that large chunk was at prices similar to where they are at currently as opposed to the highs near $120k/bitcoin. Nonetheless, it put MARA much more in the camp of Strategy ( MSTR ) and less in the that of the other bitcoin miners. They have ever since been a much more leveraged play on the direction of bitcoin, in the way that MSTR is, and are no longer "just" a bitcoin miner. In fact, MARA has the second highest amount of bitcoin on their balance sheet of any public company. MARA Makes a Different Kind of Deal For this reason, I'm not sure a deal similar to their peers would have saved MARA. They are still much more tied to the price of bitcoin and leasing out their equipment to be used for other companies' HPC needs wouldn't entirely fix their extreme tie to the price of bitcoin. There are a couple more pieces of news from MARA that have cemented the fact that they are going about this a slightly different way than most of the other miners. First, they purchased Exaion, which the transaction closed around a week ago , after being originally proposed by MARA in August of 2025. This sets them up to provide HPC infrastructure internationally. In addition to closing the Exaion deal a week ago, they just announced in their latest earnings call that they are partnering with Starwood Digital Ventures to "develop, finance and operate next generation digital infrastructure capable of meeting growing demand from enterprise, hyperscale, and AI customers." Essentially MARA is providing the data center sites with access to cheap energy access and Starwood is helping invest, as well as provide their expertise on setting up such data centers and procuring tenants. The deal is structured as a partnership, where MARA can invest up to 50% in joint venture projects. They will originally be expecting to provide around 1 gigawatt of capacity in the initial stages and up to 2.5 gigawatts over time. To me, this is similar in one way to the deals other miners have made in that they are essentially tying up an amount of their infrastructure to focus on HPC. However, the deal MARA made is very different from the others in that it is more of a partnership with Starwood, whereas the others were agreeing to supply X amount of gigawatts for Y amount of money over so many years. Those other miners have much more stability in that they know exactly what they are getting in revenues for leasing out their infrastructure. MARA is taking on a little more risk by being a joint partner in this, but they should also see much more of the upside. Both the Exaion deal and this partnership with Starwood puts MARA in position to give hyperscalers and/or other smaller companies access to GPUs to train and run their AI models on. I think this means MARA is looking to go the route of Fluidstack, CoreWeave, and Nebius ( NBIS ), giving customers access to cloud resources, and less like their mining peers that have mostly leased out their hardware to such companies as these. To me, this means MARA will be much more like an option-like play. While their peers gave themselves much more safety by leasing out their hardware in long-term deals, they also limit their upside because their revenues are more or less decided for many years. MARA however, could see very large upside if they can actually put themselves into a growth scenario like Nebius and CoreWeave, who both trade at much higher P/S multiples. It will be tough to compete with companies such as Nebius and CoreWeave, but if they can manage, it could be very lucrative for MARA. One of the things I really like about MARA is that they realize that the key to the HPC business, is their access to low-cost energy. They have been very clear about this in recent months, with their CEO Frederick Thiel stating, "we believe those who control abundant, low-cost energy will shape the future of both finance and intelligence" on a previous earnings call. I think MARA has a leg up over some of their competitors in this regard. CoreWeave and Nebius are both growing at break-neck speeds in order to try to meet the demand being seen right now. While I think this is very smart for CoreWeave and Nebius in the short-term, it may lead them to have higher costs by building data centers wherever they can with enough capacity. By focusing on the energy side of things ahead of growth at any cost, it will make sure MARA can have as cheap of access to energy as possible and should help margins much more. There still are a lot of execution risks by MARA acquiring a controlling stake in Exaion, as well as well as the partnership with Starwood. I don't really think MARA will be able to compete with the sort of growth of CoreWeave and Nebius. However, I do think they will be able to make a very profitable business based on their expertise with GPUs and driving costs down by accessing the lowest cost energy available (mostly in Texas). They have a lot of experience dealing with the foundations of HPC (GPUs and energy), with their years in the bitcoin mining business and Starwood should hopefully be able to bring them up to speed with the rest of the business. Outside of the big Starwood news, MARA posted quarterly revenues of $202M, that were actually down 5.6% Y/Y. They also posted large GAAP losses of -$4.52, but these were of course mostly due to bitcoin falling so far, so their bitcoin holdings were marked to market. This was somewhat expected, though the actual amounts are kind of eye-popping. MARA continued to drive down the cost per petahash, so it was from lower blocks won that contributed to most of the miss in revenue. Despite the misses, the market sent them about 15% higher after market, as I'm writing this. So it seems the market is positive about this news, at least as far as a knee-jerk reaction goes. We'll see how it plays out going forward Conclusion MARA is both risky due to so much of their balance sheet being bitcoin holdings and because they decided to go the partnership route instead of leasing out their infrastructure. One big positive is that they should be able to retain more control of where they use their assets, so they can still put more cycles toward bitcoin mining if prices surge back past previous highs. They also should be able to participate in more of the upside of their HPC endeavors than peers, but they will undeniably be taking on more risk than a lot of their peers. There are a wide range of possibilities still- if bitcoin bounces back and goes to new highs in the next 2-5 years, MARA is likely to benefit greatly from this due to just how much bitcoin they carry on their balance sheet. And if bitcoin has already seen its all-time highs and keeps dropping in the coming years, that too could ruin their share price, despite their plans to diversify their business. Likewise, if MARA can compete with CoreWeave and Nebius, it could expand their multiples greatly, as well as increase their revenue growth. But if they struggle to compete with such companies, they could find themselves in an even riskier position than they were as a pure-play bitcoin miner. I can't be sure at all what will happen, but I imagine it will be somewhere in between the extremes. I personally think bitcoin will still be under pressure for the next 6-12 months, but then will continue its long march higher. For this reason, I think MARA is worth the risk, but one does need to be ready to lose a considerable chunk of their investment if they choose to invest in MARA. They should be an option-like play with the ability to return several hundred percent if things go well, but also are very risky with basically the entirety of the investment at risk if things go poorly.