Report contents
Industry landscape
Where the company competes: global hashrate, local power
MARA Holdings competes in industrial-scale Bitcoin self-mining. It builds and operates data centers filled with ASIC mining rigs (chips built solely for bitcoin mining), converts electricity into hashrate (the computing power used to mine bitcoin), and earns bitcoin through block rewards and transaction fees. Its Q3 2026 shareholder letter describes a strategy to maximize the value of every megawatt and to add data center leases and GPU-as-a-service. But the base business remains mining. The precise submarket is not "crypto" in general. It is the ownership and operation of powered sites running bitcoin mining rigs, plus an early attempt to sell power and shell capacity to third-party tenants.
The value chain has three relevant links. Upstream sit power generators, grid operators, landowners, transformer and substation suppliers, and ASIC manufacturers. Downstream sit the Bitcoin protocol, mining pools, exchanges and OTC desks that convert mined coins into dollars, and, for the new business, data center and GPU customers. Bargaining power is concentrated upstream. A miner can choose among equipment vendors, but the leading suppliers set prices and delivery schedules; grid operators control interconnection and often the most valuable input, firm power. Downstream, no customer negotiates a price for block rewards. The bitcoin price is set by the market, and network difficulty adjusts to total hashrate. MARA's profit is therefore a residual: revenue per bitcoin minus power, hosting, depreciation and corporate costs. That makes low-cost power and high uptime the only plausible sources of a durable cost advantage.
Geography cuts two ways. Competition for blocks is global. A miner in Texas competes with a miner in Kazakhstan, Ethiopia or Paraguay for the same block subsidy, because the Bitcoin network does not care where the hash is produced. Competition for power, land, permits and interconnection is local. MARA's operating footprint is in US power markets; the Q3 2026 letter points to Long Ridge and a 2 GW site in Matagorda County, Texas, as examples of that local race. The distinction matters: scale in energized hashrate does not by itself confer pricing power. A large miner can spread corporate overhead and buy equipment more cheaply, but it cannot charge customers more for a bitcoin than the market price. Its cost position depends on the price it pays for electricity, the uptime of its sites and the efficiency of its fleet. MARA reported purchased energy cost of $39,674 per bitcoin at its owned sites in Q3 2026, equal to $0.04 per kWh, according to its shareholder letter. That is a useful marker of the model: a few cents per kilowatt-hour separates a profitable site from a marginal one.
Market structure is fragmented. The listed peer set includes CleanSpark, Riot Platforms, Cipher Mining, TeraWulf and Hut 8, alongside MARA and a long tail of private and hosted miners. No single company sets the bitcoin price or the network difficulty; both are determined outside any miner's control. Public hashrate share estimates move with rented capacity and are not a stable ranking. Hashrate is mobile: miners can move rigs, rent capacity or curtail when power prices spike. That mobility limits the value of installed scale. A miner with a large fleet still has to earn a return on each machine against a global difficulty level that rises as more capacity joins the network.
Barriers are mixed. Building a large, interconnected site is expensive and slow. It requires land, substations, transformers, permits and a place in an interconnection queue that can take years. Those frictions protect incumbent sites in good power markets. But entry at the margin is easier. A smaller operator can buy ASICs, lease space from a host and compete for blocks without owning generation. The technology is available from merchant suppliers, the product is homogeneous, and customers do not sign long-term contracts in the mining business. The practical result is low concentration in the mining layer and medium barriers at the site level, not a market where three or four firms control supply.
The emerging data center and GPU business has a different structure. There, MARA would compete for tenants against data center developers and power-rich sites, and contracts would be bilateral and long-term. That could raise barriers and improve revenue visibility. So far the evidence only shows an intention: the Q3 2026 letter says MARA is targeting two leases before year-end and that timing could slip into next year. Exaion signed a four-year $79 million GPU-as-a-service contract on 100% take-or-pay terms, the letter says, but that is a small signal next to the mining base. Until leases are signed, the competitive arena remains a commodity mining market with local power advantages, not a contracted infrastructure business.
That structure sets the terms for the moat question. MARA can have a cost edge at particular sites, especially where it controls generation or pays low industrial rates. It does not have the usual markers of a wide moat in its core market: no customer captivity, no proprietary standard, no regulatory franchise over bitcoin mining, and no evidence that its scale lets it set price. The exception would be a shift into contracted powered-shell leases, where location and interconnection could create local monopolies. Whether MARA can turn site-level power advantages into a durable return advantage is the open question.
Competitive advantage
Does the company have a moat? Only at the site level, and weakly
Moat rating: Weak. MARA Holdings owns things a rival cannot order from a catalog: energized sites, substations and a 2 GW campus under development in Matagorda County, Texas. It has not converted them into a return that survives a bitcoin cycle. Operating cash flow was negative in every fiscal year from 2021 through 2025 and in each of the first three quarters of 2026; free cash flow was negative in all eight of those periods. The point a reader may miss is that MARA's reported profit tracks the coins sitting on its balance sheet more than the spread between power and bitcoin, so the income statement looks strongest in quarters when mining economics matter least.
The September 2026 quarter shows it. Revenue was $0.16 billion, operating income $0.18 billion and net income $0.63 billion, the 10-Q filed November 9, 2026 shows; the gap sat below the operating line, where digital assets are marked. Three months earlier, on $0.17 billion of revenue, the company reported a net loss of $0.61 billion. Mining margins do not move that far in a quarter; coin marks do. That is why the cash flow statement, not the income statement, is the honest scorecard for whether a barrier exists here.
High profit in a single year would not settle the question either. Bitcoin's block reward is a fixed number of coins, the price is set by a global market, and difficulty rises as capacity joins the network, so a cost saving that a miner finds gets competed away through more hashrate within a few difficulty adjustments. Only a return measured against the capital tied up in sites and machines, sustained across a cycle, indicates protection. MARA's capital base has grown; its return has not.
Market position: operating scale without a defensible rank
Market position: Moderate. Revenue rose from $0.16 billion in 2021 to $0.91 billion in 2025, and MARA sits among the handful of listed miners at that scale, with CleanSpark, Riot Platforms, Cipher Mining, TeraWulf and Hut 8. Scale buys cheaper equipment, cheaper capital and a spread of sites across several power markets. It does not buy a rank that a rival must overcome: rigs are rented and moved, and the mining layer runs from a few listed companies down to single-site private operators. No published series fixes a stable share for any miner, and MARA's own position moves with how much capital it deploys each year. Moderate fits real operating scale in a market with no measured leader and no route from size to pricing.
Position stability: the footprint was bought, and the revenue fell anyway
Stability: Weak. The evidence that MARA has held ground is capacity, and capacity was financed. Financing cash flow was $1.63 billion in 2025 and $0.63 billion in the third quarter of 2026 alone, against operating cash flow of negative $0.80 billion and negative $0.17 billion. Even with that funding, total assets fell from $7.29 billion at December 31, 2025 to $5.68 billion at September 30, 2026 as the digital asset balance shrank. Revenue in the September 2026 quarter, at $0.16 billion, was down from $0.25 billion a year earlier. More machines and less revenue is what a rising network difficulty looks like from the inside. A position that has to be re-bought every time difficulty steps up is not stable in the sense that matters for a moat.
Profitability: the cash leaves in every period of the record
Profitability: Weak. The pattern across the periods covered here, taken from the company's annual filings and the September 30, 2026 10-Q (USD bn, rounded):
| USD bn | 2021 | 2022 | 2023 | 2024 | 2025 | 9M 2026 |
|---|---|---|---|---|---|---|
| Revenue | 0.16 | 0.12 | 0.39 | 0.66 | 0.91 | 0.50 |
| Operating income | -0.06 | -0.67 | 0.22 | 0.31 | -0.82 | -0.34 |
| Net income | -0.04 | -0.69 | 0.26 | 0.54 | -1.31 | -1.24 |
| Operating cash flow | -0.02 | -0.18 | -0.32 | -0.68 | -0.80 | -0.89 |
| Free cash flow | -0.73 | -0.22 | -0.34 | -0.93 | -1.21 | -1.16 |
Two years of positive net income, 2023 and 2024, were also years when the bitcoin price rose; neither repeated. Average ROE for 2021-2025 was about negative 25%, and leverage is a live factor rather than a rounding item, with total liabilities at 53% of assets on September 30, 2026.
An after-tax return on invested capital estimate, using operating income taxed at the year's effective rate and invested capital as equity plus debt minus cash, tells the same story. For fiscal 2025, NOPAT of roughly negative $0.82 billion on invested capital of about $6.52 billion gives an estimated ROIC near negative 13%. For fiscal 2024, NOPAT of about $0.27 billion on $6.19 billion gives roughly 4%, estimated. One year of single-digit returns beside three years of double-digit losses sits below the 6%-8% band that usually signals no advantage, let alone the 15%-25% that marks one.
Reported gross margin is not a useful read on unit economics here: it ran from negative 93% in the first quarter of 2026 to positive 72% in the second, swings that point to digital-asset accounting inside cost of revenue. The cash figures cut through it.
Overall rating: weak, because the cheapest power has not produced a cash-generating year
Overall rating: Weak. Profitability carries the most weight and it is Weak, which caps the overall call regardless of the Moderate market-position score. The decisive facts are simple: negative operating cash flow in all eight periods covered, an estimated return on capital that was about 4% in the best recent year, and revenue falling year over year in the latest quarter while the asset base and debt grew. A cost advantage that never shows up as a year of positive cash generation has not yet been demonstrated.
Where the advantage comes from and how long it can last: cheap power at a few sites, and a lease business not yet signed
Two documents carry the operating detail. The 10-Q for the quarter ended September 30, 2026 supplies the balance sheet and cash flow; the quarterly shareholder letter supplies the operating numbers that decide whether an edge exists, chiefly the power cost per bitcoin at sites MARA owns. The single metric that tests this moat is that figure: what MARA pays in electricity to produce one bitcoin, tracked against the price it receives for it. MARA reported purchased energy cost of $39,674 per bitcoin at its owned sites in the third quarter of 2026, equal to about $0.04 per kWh, per the letter.
Supply side: cheap power is a real edge, and a local one
Supply-side advantage: Moderate. Electricity is the largest variable cost in mining and the product is identical whoever produces it, so a site that buys power cheaply earns a wider spread on every coin than a site that does not. That is a genuine cost advantage, and it is structural where MARA controls the interconnection or holds a contract that a rival cannot obtain at the same price: substations, transformers and permits take years, and the queue for firm power in any given market is finite. The protection is local, though, and thin in one direction. Power is a commodity that other operators can contract for, and the rate MARA pays can be reset by a utility rate case, by a change in curtailment payments, or by data centers bidding for the same capacity. The arithmetic shows how much rides on cents. At a 1 GW site running flat out, one cent per kWh is about $87.6 million a year (1,000,000 kW × 8,760 hours × $0.01), estimated; three cents is roughly $263 million, about 40% of MARA's annualized third-quarter revenue of roughly $0.64 billion. Whether $0.04 is cheap depends on the local tariff, and the same number is less impressive for curtailed off-peak supply than for firm power. What would break the advantage: a reset in the local power price, or a curtailment program that pays less than the coins the idled rigs would have mined.
Demand side: there is no customer to hold
Demand-side advantage: None. Mining has no customer captivity because it has almost no customers in the ordinary sense. Coins are sold at the market price; a mining pool pays the same for a share of hashrate regardless of who produces it. Nothing renews, nothing needs retraining, nothing integrates with a buyer's systems, and a rival's discount cannot take share because there is no price to discount. That absence cuts both ways. It means the only thing that wins blocks is more hashrate at lower cost, and nothing stops a funded entrant from buying that. The one place captivity could appear is the data center and GPU leasing business, where tenants sign multi-year contracts with penalties for leaving. Exaion's four-year, $79 million GPU contract on 100% take-or-pay terms, disclosed in the letter, is a real obligation of that kind, but about $20 million a year against roughly $0.64 billion of annualized mining revenue is a little over 3% of the total. So far the evidence only shows a toehold.
Scale: interconnection is the scarce asset, not the rigs
Scale advantage: Moderate. Fixed costs in mining sit in site development, substations and corporate overhead; the rigs themselves are bought at market from a small group of suppliers whose prices and delivery schedules are broadly available to any buyer with capital. That limits what size alone can achieve. Where scale does bite is in assembling large blocks of firm power and in spreading corporate and engineering costs across more megawatts, and the Matagorda campus is the clearest expression of it: a 2 GW project is a different proposition from a 50 MW site when the bottleneck is the interconnection queue. Without customer captivity, though, scale can only lower unit costs; it cannot lift price. And the same queue that keeps new miners out is open to the AI data center developers now competing for the identical resource, which raises MARA's opportunity cost per megawatt at the same time it creates the lease option. What would break the scale advantage: a market where interconnection stops being scarce, or one where a rival with a bigger balance sheet reserves the available capacity first.
Durability: difficulty, the halving and a rival buyer for the power
The clock on all of this is set outside the company. Network difficulty rises as capacity joins, and the next halving, scheduled by the protocol for 2028, cuts the block subsidy again, doubling the cost per coin at any given power price unless bitcoin rises or the fleet gets more efficient. A miner that cannot cut its all-in cost per coin in step with those two forces sees its spread compress regardless of how good its sites are. The substitute risk is unusual because it comes from the input side: if AI and cloud tenants will pay more per megawatt than mining earns, MARA's best use of its best sites stops being mining, which would either shrink the mining base or convert part of the company into a landlord with contracted revenue. Access to capital is a false moat in the meantime. Financing supplied $1.63 billion in 2025 while operations consumed $0.80 billion, and equity and debt markets are open to every listed miner, so raising money to buy machines does not keep anyone out.
Questions the record leaves open
- Does power cost per bitcoin stay near the third-quarter level as Matagorda comes online, or climb with difficulty? The first would show a site-level edge; the second would show scale of operations and little more.
- Do the two data center leases MARA said it was targeting for late 2026 get signed, and at what rent per megawatt relative to what mining earns on the same capacity? The letter allows that timing could slip into 2027.
- What bitcoin price does the all-in cost per coin require once the 2028 subsidy cut lands, and would the current power contracts still clear it?
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Comparison and summaryHow MARA Holdings, Inc. compares with its closest rivals, and our moat rating
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on MARA Holdings, Inc.
MARA Holdings, Inc.
Business, moat and key questions
Open company page