Marathon Digital Holdings, Inc. (MARA) Earnings

Management highlights

Strategic Transition - MARA is executing a strategic transition from a pure-play Bitcoin miner to a digital infrastructure company focused on controlling and monetizing low-cost power for AI and compute, leveraging Bitcoin mining as an immediate revenue foundation that retains flexibility to redirect capacity to AI as opportunities mature - The core strategic thesis is that available, connected energy is the primary bottleneck to AI compute growth, and controlling power provides a structural competitive advantage - Two complementary AI market pathways are being pursued: large-scale hyperscale co-location via the Starwood joint venture, and sovereign/private/enterprise cloud AI via the majority-owned Exxon segment Strategic Transaction Progress - Closed acquisition of a majority interest in Exxon, targeting the sovereign and private cloud AI market segment driven by growing demand for data sovereignty, jurisdictional compliance, and independent compute capacity, with active discussions ongoing in France, Brazil, and Saudi Arabia - Moved the Starwood joint venture from announcement to execution, with permitting and site preparation advanced across 90% of MARA's existing owned and operated sites, and active tenant discussions underway with multiple hyperscalers - Shortly after quarter end, announced a definitive agreement to acquire Long Ridge Energy and Power from FTAI Infrastructure, an operational existing power and land asset adjacent to MARA's Hannibal Bitcoin mining site - Retired 30% of outstanding convertible debt at a discount, reducing potential fully diluted share count by approximately 46 million shares (9%) and strengthening balance sheet flexibility - Implemented a 15% workforce reduction to realign the organization for the digital infrastructure strategy, generating $12 million in annualized cost savings Long Ridge Acquisition Highlights - Long Ridge adds 1,600 acres of contiguous land and an existing 505 megawatt nameplate efficient combined cycle gas turbine, with a path to scale total power capacity to over 1 gigawatt and 600 gross megawatts of AI/IT load - The asset generated $144 million annualized adjusted EBITDA in H2 2025 with 76% contracted capacity, providing immediate stable cash flow upon closing - Management estimates building a comparable asset from scratch would require $2 to $3 billion in capital and 7 to 10 years of development time - Initial 200 megawatts of AI build is planned to start construction in H1 2027, with initial capacity coming online in mid-2028 - Existing Bitcoin mining operations at Hannibal will continue uninterrupted until capacity is needed for AI campus development Starwood Joint Venture Benefits - Starwood is an experienced institutional data center developer with a 7 gigawatt global track record, which accelerates tenant evaluation and lease signing and provides confidence to prospective tenants around meeting timelines and technical requirements - Starwood provides captive development and EPC capabilities, eliminating the need for MARA to source and manage third-party contractors - The joint venture structure is capital efficient: MARA receives equity credit for contributed sites based on pre-agreed values of power, land, and interconnection, requiring little to no incremental equity from MARA for early projects, with incremental growth capital split proportionally for larger developments. An illustrative 200 megawatt project could generate $50 to $100 million in annualized stabilized net cash flow for MARA with minimal incremental equity Financial Performance - Total Q1 2026 revenue was $174.6 million, down from $213.9 million in Q1 2025, primarily due to an 18% drop in average Bitcoin price that reduced revenue by $33.1 million - Reported net loss was $1.3 billion ($3.31 diluted loss per share), up from $533.4 million ($1.55 per share) in Q1 2025, with ~$1 billion of the loss driven by unrealized mark-to-market fair value adjustments on Bitcoin holdings triggered by Q1 price declines - Adjusted EBITDA was negative $1 billion in Q1 2026, compared to negative $483.6 million in the prior year quarter, also dominated by Bitcoin mark-to-market changes - G&A expense (excluding stock-based compensation) was $57.7 million, up from $36.9 million year-over-year, including $11 million in acquisition and integration costs; after restructuring, the recurring quarterly G&A run rate (excluding acquisition costs and stock comp) is expected to decline below Q1 levels - MARA held 35,303 Bitcoin at quarter end, with 28% loaned or pledged as collateral generating $6.4 million in Q1 interest income; $1.5 billion in Bitcoin was sold during the quarter to fund debt reduction, with no use of the at-the-market equity offering program since Q3 2025

Guidance

- Management expects to sign multiple tenant leases for AI capacity by the end of 2026, with contracted megawatt disclosures to follow as the project pipeline converts - Initial 200 megawatts of AI build-out at the Long Ridge campus is planned to start construction in the first half of 2027, with initial capacity coming online in mid-2028; the full 600 megawatts of AI capacity is expected to come online incrementally over 18 to 24+ months after closing - Long-term, management expects the customer mix for AI capacity to be heavily weighted toward hyperscalers in the near term (roughly 90% hyperscaler / 10% enterprise), shifting to approximately 60% hyperscaler / 40% enterprise over time as private cloud demand grows - Management expects recurring quarterly G&A run rate (excluding stock-based compensation and non-recurring acquisition/integration costs) to trend below Q1 2026 levels as restructuring savings are realized - Inference demand is expected to grow dramatically over the coming years alongside continued growth in AI model training demand, driven by the expansion of agentic AI technology that increases token consumption by orders of magnitude

Segment performance

MARA's primary segment in Q1 2026 is Bitcoin mining, which forms the operational foundation for the company's new AI infrastructure strategy. Bitcoin mining generated total Q1 2026 revenue of approximately $170.9 million (accounting for 97.9% of total company revenue of $174.6 million, with the remaining 2.1% coming from other digital asset hosting services). Key Bitcoin mining operational and financial metrics: record energized hash rate of 72.2 exahash per second (up 33% year-over-year), 2,247 total Bitcoin mined (25 per day, 39 fewer than Q1 2025 due to higher network difficulty partially offset by higher hash rate), market share of available mining rewards reached 5.5% (up from 4.8% in Q4 2025), cost per kilowatt hour at owned sites was $0.04 (among the most competitive at scale in the sector), cost per petahash per day improved 3% year-over-year to $27.6 (down 42% over 11 quarters), energy cost per mined Bitcoin was $40,047 (up from $35,728 in Q1 2025 due to higher network difficulty). New AI digital infrastructure segments are in development, with no material revenue contribution in Q1 2026. These segments include large-scale hyperscale AI capacity developed via the Starwood joint venture, and sovereign/private/enterprise AI capacity via the majority-owned Exxon segment.

Risks & headwinds

- Forward-looking statements regarding AI development timelines, tenant lease signings, and financial performance are subject to inherent risks and uncertainties, and actual results may differ materially from expectations, as noted in the company's SEC filings - Bitcoin price volatility creates significant swings in reported net income and adjusted EBITDA via mark-to-market adjustments; every $10,000 change in Bitcoin price impacts the fair value of MARA's holdings by approximately $350 million - AI capacity development depends on successful grid interconnection approvals and permitting, which can face unanticipated delays - Closing of the Long Ridge acquisition is subject to regulatory approval, debt holder consent for the existing secured notes, and satisfactory financing, with market conditions impacting the availability and cost of funding - Conversion of existing Bitcoin mining capacity to AI use requires tenant demand to materialize on expected timelines, and the transition will take place gradually over multiple years - Competitive demand for AI infrastructure capacity could lead to pricing pressures or slower-than-expected lease signing - The strategic transition to digital infrastructure requires new organizational capabilities, and misalignment of skills could impact execution

Analyst Q&A

• Q: How should investors think about MARA's broader strategy of commercializing existing sites versus acquiring new opportunistic capacity like Longridge? Is Longridge a one-off, or will this mixed approach continue?

  A: Longridge was not a random one-off; the deal had been in planning since MARA originally acquired the adjacent Hannibal site, and it aligned with the original plan to build a large premier AI campus at Hannibal. Going forward, MARA will continue pursuing a mixed approach: adding small tuck-in sites that can be developed for smaller-scale AI inference use cases (which fit well with MARA's existing modular development experience from Bitcoin mining) and continuing to pursue larger land and power opportunities to develop big campuses with the Starwood joint venture. This dual approach allows MARA to serve both large hyperscale and smaller enterprise/inference demand.

• Q: What is the timeframe for developing the 600 megawatt AI campus at Longridge, and what is the status of additional grid connection?

  A: Behind-the-meter capacity expansion is already in process and will come online first. The first 200 megawatt AI facility will take 18 to 24 months to come online, with an additional 200 megawatts of behind-the-meter capacity available by that time. The remaining 200 megawatts from grid interconnection expansion will come online shortly after that. The priority is getting the first tenant site operational, with additional capacity already queued and ready to develop.

• Q: Why is management confident the first Starwood joint venture tenant lease will be signed this year, given the competitive AI infrastructure market?

  A: Strong demand from prospective tenants is driving confidence: multiple prospective tenants are competing for capacity across MARA's sites, and demand for AI compute is not decreasing. Model providers are directly limited in their growth by available compute capacity, as the explosion of agentic AI has led to far faster-than-expected growth in token consumption that has strained existing capacity. Many enterprises are also finding public cloud AI costs uneconomical and are seeking private on-prem or near-prem capacity, further increasing demand for available powered sites like MARA's.

• Q: What explains the recent increase in G&A, and what is the expected path for future costs after the recent workforce reduction?

  A: The Q1 G&A increase includes $11 million in non-recurring acquisition and integration costs, plus investments in new skills required for the transition from a pure Bitcoin miner to a digital infrastructure company. The recent 15% workforce cut eliminated roles that are no longer needed for the new strategy. The recurring quarterly G&A run rate (excluding non-recurring transaction costs and stock-based compensation) is expected to decline below Q1 levels over time as restructuring savings are realized. The transition is gradual, as existing Bitcoin mining operations continue running while AI sites are developed, so operational staff will be reduced incrementally.