Bitcoin Miners Report Over $1.8B in Combined Losses as AI Pivot Accelerates
Three of North America’s largest publicly traded Bitcoin miners reported a combined net loss exceeding $1.8 billion for the quarter ended March 31, 2026, as a sharp decline in Bitcoin’s price collided with structurally higher mining difficulty to compress revenues across the sector. MARA Holdings, CleanSpark, and Keel, the company formerly known as Bitfarms, each posted results that underscore a deepening incompatibility between the pure-play mining model and the financial expectations of public equity markets. The common thread running through all three sets of results is not operational failure but rather the consequences of holding large Bitcoin treasury positions through a quarter in which the asset fell approximately 23 to 25 percent.
The Numbers in Detail: Revenue Compression and Mark-to-Market Damage
MARA Holdings reported Q1 2026 revenue of $174.6 million, an 18 percent decline year-on-year and a miss against Wall Street’s consensus estimate of $192.7 million. Its net loss for the quarter reached $1.3 billion, more than double the $533.4 million loss recorded in the same period a year earlier. Earnings per share came in at a loss of $3.31, against an analyst estimate of a loss of $2.20. The company attributed the bulk of the loss to unrealized markdowns on its 38,689 Bitcoin treasury, as BTC moved from roughly $87,000 to approximately $67,000 during the quarter. To manage liquidity, MARA sold more than 15,100 Bitcoin worth approximately $1.1 billion in the final week of March alone, using $1.0 billion of the proceeds to retire convertible notes.
CleanSpark’s results covered its second fiscal quarter, also ending March 31. The company reported a net loss of $378.3 million, with Bitcoin mining revenue of $136.4 million representing a 25 percent decrease from the $181.7 million it recorded in the comparable prior-year period. A $224.1 million non-cash loss on Bitcoin fair value accounted for the largest single component of the decline, even as CleanSpark continued to expand its hashrate and power capacity during the period. The operational trajectory and the reported financial outcome thus point in opposite directions, a divergence that reflects the accounting treatment of Bitcoin holdings under fair value rules rather than any deterioration in the company’s mining infrastructure.
Keel, which rebranded from Bitfarms as part of its strategic repositioning, reported a Q1 net loss of $145 million on revenue of just $37 million. The revenue figure is notably thin relative to its larger peers, reflecting both the scale gap and the disruption costs associated with pivoting a business model mid-cycle. Despite those numbers, Keel’s shares rose following the announcement, alongside a $533 million liquidity update, suggesting that equity markets are assigning meaningful forward value to the AI infrastructure thesis rather than penalising current-period losses. That reaction is worth examining carefully, because it implies investors are pricing in a successful transition that has not yet produced material revenue.
Structural Forces Behind the Losses
The sector’s difficulties do not stem from a single quarter’s price weakness. Bitcoin is currently trading more than 35 percent below its all-time high of $126,080, which was set in 2025, and mining difficulty has risen nearly 30 percent over the past year. At time of writing, the network hash rate sits at approximately 1,086.6 exahashes per second, a figure that illustrates the continued investment in mining capacity even as per-block revenues have contracted. The combination of a lower Bitcoin price and higher difficulty means each unit of hash rate generates materially less dollar-denominated revenue than it did twelve months ago, and that squeeze is arithmetic rather than cyclical in the short term.
The April 2024 halving, which cut the block subsidy from 6.25 BTC to 3.125 BTC, established the baseline for this pressure. With approximately 100,950 blocks remaining until the next halving at time of writing, miners have roughly two years before another 50 percent reduction in subsidy income arrives. That timeline is relevant context for the AI pivot: companies that can redirect power capacity toward AI and high-performance computing workloads within that window will enter the next halving cycle with a fundamentally different revenue structure. Those that cannot will face a second compression event on top of the current one.
It is also worth separating two distinct sources of loss in these results. The first is operating cash flow, which reflects actual mining economics. The second is non-cash fair value adjustments on Bitcoin holdings, which are driven entirely by price movements and do not affect liquidity. CleanSpark’s $224.1 million fair value loss and MARA’s treasury-related markdowns fall into the second category. This distinction matters for assessing the underlying health of operations, but it does not diminish the headline impact on reported earnings, and it does not change the fundamental challenge of generating adequate returns from mining alone in the current environment. The accounting framework, adopted following changes to FASB rules, requires companies to mark Bitcoin holdings to market each quarter, amplifying both reported gains in rising markets and reported losses in falling ones.
The AI Pivot: Strategy Versus Execution Risk
All three companies have announced or are pursuing transitions toward AI and high-performance computing infrastructure, and the market appears to be rewarding the articulation of that strategy even before revenues materialise. MARA’s approach is the most capital-intensive of the three. The company agreed in late April to acquire Long Ridge Energy and Power from FTAI Infrastructure for $1.5 billion, a gas-fired power plant and data center that management says could eventually support 600 megawatts of AI computing capacity. MARA has also partnered with Starwood Capital to develop data centers targeting one gigawatt of computing capacity, and its statement that around 90 percent of its non-hosted mining capacity could be redeployed for AI and IT compute suggests the company views the infrastructure transition as largely achievable without greenfield construction.
MARA’s own language is instructive here. The company described Bitcoin mining as its “operational foundation” while simultaneously stating it has no future plans to purchase additional Bitcoin mining hardware. That pairing communicates something precise: mining continues to generate cash today, but capital allocation has shifted decisively toward AI infrastructure. The company’s co-location strategy, placing new AI infrastructure adjacent to existing mining sites, is designed to preserve optionality rather than force a binary choice between the two revenue streams. As MARA put it, the approach “creates flexibility: we can generate revenue today through Bitcoin mining while preserving the option to redirect power toward AI and critical IT loads as those opportunities mature on the same sites.”
Keel’s pivot is earlier-stage and more financially precarious given its $37 million quarterly revenue base, but the market’s positive share price reaction to a $145 million loss suggests the investment community is treating the $533 million liquidity update as sufficient runway to execute the transition. CleanSpark has been more measured in its public communications about AI, continuing to emphasise hashrate expansion while acknowledging the broader sector dynamics. Whether that positioning reflects disciplined focus or slower strategic adaptation is a legitimate question that the next two to three quarters will clarify.
Beyond these three companies, the broader direction of the sector is being set by more aggressive movers. IREN has signed a five-year, $3.4 billion AI cloud contract with Nvidia to deploy up to five gigawatts of AI infrastructure, a transaction that places it in a materially different category from miners still in the planning phase of their transitions. HIVE Digital Technologies has committed capital to a 50-megawatt computing facility supported by high-speed fiber infrastructure. These moves collectively indicate that the conversion of stranded power assets into AI compute is no longer a strategic hypothesis; it is an active capital deployment theme across the sector. The parallel pressures on Bitcoin treasury holders like Strategy, which reported a $12.54 billion Q1 loss driven by a $14.46 billion unrealized markdown on its holdings, further illustrate how mark-to-market accounting has become a defining feature of this earnings season for Bitcoin-exposed balance sheets.
Who Benefits and Who Bears the Cost of This Transition
The analysis here points in a clear direction. Companies that have already secured long-term AI compute contracts with creditworthy counterparties, as IREN has done with Nvidia, will emerge from this period with diversified, contracted revenue streams that are structurally independent of Bitcoin price movements. They will also benefit from a lower-cost basis on their power infrastructure, having acquired it during the mining era when data center developers were not yet competing aggressively for the same sites. That cost advantage is durable and is not easily replicated by hyperscalers building new facilities from scratch.
MARA occupies a more complex middle position. Its Long Ridge acquisition and Starwood partnership represent serious capital commitments, and the company has the balance sheet scale to absorb transitional losses. However, a $1.3 billion quarterly loss combined with an 18 percent revenue decline and a MARA stock that has fallen 16 percent over the past twelve months, dropping from the largest Bitcoin miner by market cap to seventh place, signals that execution risk is real and that the market is not yet fully convinced. Selling more than $1.1 billion of Bitcoin in a single week to service debt is a liquidity management action, not a sign of strategic strength, even if it was the rational move under the circumstances.
The companies most exposed to downside from here are those with high mining cost structures, limited power capacity to repurpose, and insufficient liquidity to fund an infrastructure transition before the next halving. Pure-play miners that have not yet begun the pivot toward AI or HPC will face a compounding problem: declining Bitcoin revenues, rising difficulty, and an equity market that is increasingly discounting mining-only business models. The sector-wide revenue compression documented across these Q1 results is not a temporary aberration. It reflects a structural shift in the economics of proof-of-work mining that was set in motion by the April 2024 halving and will intensify with the next one.
Network activity metrics at time of writing add a layer of context that is often absent from mining company earnings discussions. Active addresses over the past 24 hours stand at 457,665, a figure that sits below the highs recorded during the bull phase of 2024 and 2025. A less active network, combined with a lower Bitcoin price, reduces total fee revenue available to miners, compounding the subsidy reduction from the halving. The structural case for diversification away from mining is therefore stronger than the headline loss figures alone suggest. Investors pricing Keel’s shares higher on a $145 million loss and a $533 million liquidity update are, in effect, making a judgment that the terminal value of AI infrastructure assets exceeds the terminal value of a mining-only operation at current difficulty and price levels. On the available evidence, that judgment is correct.
