MARA Q1 2026 AI Spend and Capex Strategy Revealed

Q1 2026: MARA’s AI spend and capex plans—what the company disclosed

Based on MARA’s Q1 2026 earnings call excerpts, the company does not provide a single consolidated dollar figure for “AI spend” or “total capex” in Q1 2026 specifically. Instead, management frames Q1 2026 as execution on a capital allocation strategy and provides time-phased capacity build-out plans (including what will likely require capex later), with the most explicit “spend timing” tied to the Long Ridge expansion.

1) AI build-out timing (most concrete “spend plan” disclosed)

Management described the initial AI build-out behind the Long Ridge asset as:

• Initial 200 MW of AI build-out, with construction beginning around the first half of 2027 and initial capacity coming online in mid-2028. ¹²
  Implication: This is the clearest indicator of when major incremental AI-related capital spending would likely occur, even though the Long Ridge acquisition itself occurred/was announced in Q1 2026. ¹²

They also described the longer-term scaling path:

• Long Ridge provides a path to scale to 600 gross MW of AI and critical IT load over time. ¹²
  Implication: Additional capex beyond the first 200 MW would be expected as that ramp occurs, but management does not quantify capex by tranche in the excerpts. ¹²

2) Q1 2026 was about financing, acquisitions, and partnership execution—not near-term AI capex outlays

Management characterizes Q1 2026 as “redefining” execution “across multiple fronts,” including closing an acquisition, retiring convertible debt, moving a JV “from announcement to execution,” and announcing Long Ridge shortly after quarter-end. ¹
They also explicitly connect this strategic execution to converting energy infrastructure into AI/critical IT. ¹²

Key point: the excerpts emphasize that these steps improve the platform and power/control position, while the AI campus build-out is largely scheduled for 2027–2028. ¹²

3) Long Ridge specifics that matter for capex planning

The company provided substantial project-level detail that typically drives future capex needs:

• Long Ridge is described as an operational site with “immediate access” to infrastructure, interconnection, land, and power needed to scale. ³⁴⁵
• The centerpiece is an approximately 505 MW nameplate combined cycle gas turbine, which generated $144 million of annualized adjusted EBITDA in 2H 2025 with 76% contracted capacity. ⁵
• MARA stated it is not reducing Long Ridge’s current power supply into PJM, and instead will develop compute behind the meter while pairing demand with incremental generation over time. ⁶
  Implication: AI-related capex would likely be deployed to build “compute capacity behind the meter” first, while generation expansion capex follows approvals and over time. ⁶⁷³

They also stated:

• They have submitted plans to augment the Hannibal interconnect, and will move quickly post-close to further expand power capacity at the power plant. ⁵
  Implication: This suggests near-term engineering/permitting and some expansion-related spending may begin around closing, even though major AI construction is guided to 2027. ⁵

4) How strategic acquisitions and capital expenditures relate (Long Ridge vs. Starwood JV)

Long Ridge (acquisition-driven capex runway)

Management frames Long Ridge as a “land and power acquisition” to develop a premier compute campus. ¹
They also quantify the operational scale-up impact:

• Long Ridge “gives us over 1 gigawatt of total potential capacity” and a path to 600 gross MW of AI/critical IT load over time. ³
• Transaction expected to increase owned/operated energized capacity from ~1.3 GW to ~2.2 GW by closing (and roughly 2.4 GW including expansion capacity). ³

They further provide a concrete near-term “capex-adjacent” timeline for AI infrastructure construction (200 MW starting 1H 2027). ³

Starwood JV (capacity monetization model, reducing incremental investment)

MARA’s Starwood partnership is described as capital-efficient for turning its powered land portfolio into contracted digital infrastructure:

• Starwood is said to provide a “capital-efficient path” to convert MARA’s power portfolio into AI infrastructure ownership. ²
• MARA also explained that in the JV, MARA contributes a site once a tenant is signed, receiving credit based on predetermined attributes (power/land/interconnection/development). ²
• Management expects to “sign multiple tenant leases by year-end,” and will disclose “contracted megawatts” as the pipeline converts. ⁶

Implication: While the excerpt doesn’t give capex dollars for Q1, the JV structure is specifically intended to avoid MARA having to invest incremental dollars for every incremental site/tenant, relative to a direct development model. ^28910111267

5) “AI spend” in Q1 2026: what can and cannot be concluded from the excerpts

From the provided excerpts, we can state with confidence:

• The earliest explicit AI construction schedule they gave is 200 MW starting 1H 2027 with mid-2028 initial capacity online—so the larger AI capex wave is not scheduled to occur in Q1 2026. ³
• Q1 2026 execution focused on balance-sheet strengthening, JV progression, and acquisition actions that set up future AI capex. ¹²

However, the excerpts do not disclose:

• A specific Q1 2026 AI spend dollar amount (e.g., “we spent $X on AI capex in Q1 2026”).
• A specific Q1 2026 capex total or an AI/capex allocation breakdown by category (data center build, power plant upgrades, interconnection, etc.).

6) Capital structure / funding plan connected to capex timing (Long Ridge)

MARA also outlined how the Long Ridge acquisition is expected to be funded, which informs the company’s ability to execute capex later:

• Long Ridge has a $400 million term loan expected to be repaid at closing, and MARA is conducting a consent solicitation for Long Ridge’s $600 million secured notes to remain in place. ¹²
• The $115 million Can-Am facility is expected to remain in place. ¹²
• Pro forma total debt at Long Ridge expected to be ~$900 million (down from $1.1 billion previously), with ~$185 million of tack-on secured notes expected. ¹²
• MARA expects to fund remaining consideration via cash on hand, borrowings collateralized by Bitcoin, and potentially Bitcoin sale proceeds, and has a $785 million Barclays commitment letter backed by a bridge loan if needed. ¹²

Implication: This financing plan supports acquiring the platform quickly while allowing capex deployment later for AI build-out consistent with the 2027–2028 schedule. ¹²⁶⁷³

---

Bottom line answer to the question

• AI spend/capex timing: MARA’s most explicit AI infrastructure spending timeline tied to Long Ridge is an initial 200 MW AI build-out with construction starting around 1H 2027 and initial capacity in mid-2028. ³
• Q1 2026 focus: Q1 2026 was primarily about platform setup (Long Ridge acquisition execution/announcement context, Starwood JV execution, and balance-sheet strengthening), not immediate large AI construction spend. ¹²
• Strategic acquisitions (Long Ridge): Long Ridge is positioned as a land-and-power acquisition for a premier compute campus, backed by an operational power plant (505 MW nameplate) generating annualized adjusted EBITDA of $144 million in 2H 2025 on 76% contracted capacity, enabling rapid ramp to AI loads over time. ¹³⁵
• Long Ridge capex pathway and power approach: MARA says it does not intend to reduce Long Ridge’s PJM power supply; instead it will add compute “behind the meter” while pairing demand with incremental generation over time, supported by planned interconnection/power plant augmentation post-close. ⁶⁵
• Long Ridge vs. Starwood partnerships: Starwood JV is described as a capital-efficient engine to monetize powered land via tenant-signed site contributions and predetermined crediting, designed to reduce MARA’s need for incremental investment per expansion opportunity; Long Ridge is the differentiated “land and power” anchor campus that expands the AI-capable platform. ²⁶

If you want, I can quantify the implied capex magnitude using typical MW-per-dollar build assumptions—but that would require external engineering/market cost inputs, which are not provided in the excerpts.