MARA (MARA) Q2 2026 earnings review
AI Pivot Takes Center Stage While Bitcoin Volatility Batters the Bottom Line
MARA's Q2 2026 results reflect a company in structural transition. Revenue decelerated, falling 27% YoY to $174.9 million, primarily driven by a 28% decrease in Bitcoin's average price. The bottom line reversed violently: Net Income plunged from an $808.2 million profit in Q2 2025 to a $611.3 million loss this quarter. However, this is largely an accounting artifact, driven by a $343.0 million non-cash mark-to-market loss on Bitcoin holdings. Operationally, the company is stable: energized hashrate grew 22% YoY, and fleet efficiency is improving. But management's narrative has decisively shifted away from pure mining. With the pending Long Ridge acquisition and a new 2 GW site in Matagorda, Texas, MARA is hoarding scarce power assets to feed the AI hyperscaler boom. Investors must now weigh the near-term financial messiness of this transition against the long-term value of a projected 4.8 GW power portfolio.
🐂 Bull Case
MARA is securing the defining bottleneck of the AI era: energized power. The Matagorda County agreement (2 GW) and Long Ridge acquisition push the potential portfolio up to 4.8 GW, establishing MARA as a premier digital infrastructure landlord.
Management is effectively utilizing its 35,577 BTC treasury. By securing $600M in Bitcoin-backed credit facilities at a 7.56% cost of debt, MARA is funding the Long Ridge acquisition without diluting equity shareholders, preserving upside while activating idle assets.
🐻 Bear Case
Because MARA does not hedge its Bitcoin holdings, its income statement is held hostage by crypto market swings. A 45% YoY decline in BTC price resulted in a $1.5 billion YoY negative swing in fair value adjustments, masking underlying operational performance.
General and Administrative expenses (excluding stock-based compensation) are accelerating, up 73% YoY to $69.5 million. This includes $15.4M in acquisition/integration costs and a $10.2M litigation settlement, highlighting the expensive friction of pivoting from mining to AI data centers.
⚖️ Verdict: ⚪
Neutral. The strategic pivot to AI infrastructure is intellectually sound given the rising difficulty of Bitcoin mining. However, the financial translation of this pivot is heavily obscured by BTC price volatility, rising administrative overhead, and the long lead times required to sign and build out AI tenant leases.
Key Themes
Network Difficulty Squeezing Mining Margins
Despite an accelerating 22% YoY increase in energized hashrate to 70.3 EH/s, actual Bitcoin production remained relatively stable, growing only 3% to 2,422 BTC. This highlights the relentless rise in global network difficulty. Consequently, the purchased energy cost per BTC produced at owned sites increased from $33,735 to $38,690. This dynamic fundamentally validates management's strategy to pivot megawatts toward higher-margin AI workloads.
Third-Party Hosting: A Persistent Drag
While owned-site metrics are improving (cost per kWh held at $0.04), MARA's legacy third-party hosting remains a financial anchor. Third-party hosting and other energy costs were flat YoY at $69.2 million. Management confirmed these arrangements will fully expire by Q1 2028. Until then, they dilute the margin benefits of MARA's vertically integrated model.
Sovereign AI as a Differentiated Growth Vector
The Exaion platform, operating within the EU regulatory framework, provides MARA with a unique moat in the 'Sovereign AI' space. As enterprise data governance tightens, capturing workloads that cannot reside on public US hyperscale clouds (due to the CLOUD Act) offers a higher-margin, sticky revenue stream. Exaion's selection for the EU-backed AION Consortium validates this specific technological capability.
Other KPIs
Stable/Improving. This operational efficiency metric improved 4% YoY from $28.7 in Q2 2025, and marks a 27% improvement over the past nine quarters. This proves that while macroeconomic BTC prices are out of MARA's control, their ground-level engineering and site management continue to scale efficiently.
Reversing. Down from a positive $1.2 billion in Q2 2025. This metric is heavily polluted by the MTM digital asset rules. A more instructive view of the core business cash generation is obscured until MARA begins reporting normalized NOI from its Starwood JV AI deployments.
Guidance
Accelerating phase-out. The most significant high-cost hosting arrangements will begin expiring in Q3 2027, culminating in complete elimination by Q1 2028. This will mechanically drive down MARA's blended cost per kWh and improve overall gross margins.
Accelerating. Between the existing Hannibal campus, the pending Long Ridge acquisition, and the newly announced 2 GW Matagorda County site, MARA is guiding toward one of the largest powered land portfolios in the digital infrastructure industry. Execution and interconnect approvals remain the primary risks to achieving this.
Key Questions
Starwood JV Tenant Timeline
With the Long Ridge acquisition progressing and Matagorda added to the pipeline, when exactly do you expect to sign the first binding hyperscaler lease through the Starwood JV, and what does the timeline to initial revenue recognition look like post-signing?
G&A Run-Rate Normalization
G&A excluding stock-based compensation jumped 73% YoY to nearly $70 million. With $12 million in annualized savings expected from Q1's restructuring, what is the normalized quarterly G&A run-rate we should model for H2 2026 once acquisition noise clears?
Funding Matagorda CapEx
You successfully utilized BTC-backed debt for the Long Ridge deal. Given the massive 2 GW scope of the Matagorda County site, will you continue relying on BTC credit facilities to fund its development, or will the Starwood JV shoulder the majority of the capital burden?