MARA (MARA) H.C. Wainwright 28th Annual Global Investment Conference summary

AI infrastructure opportunity

• Hyperscalers are expected to spend over $1 trillion on AI next year; Bitcoin miners have secured over $160 billion of AI colocation deals, and panelists cited over 14 GW of combined power capacity in view.

• Miners’ energized power and land offer a faster route to AI compute; converting assets can materially increase revenue and asset value per megawatt versus mining infrastructure.

• Demand spans hyperscalers, frontier model providers, neoclouds and enterprise inference; lease sizes range from 50–150 MW to 400–500 MW, with large sites supporting multiple tenants.

Lease pipeline and business models

• Two leases are targeted by year-end; activity is expected across small and large sites, with pricing and lease structure tailored to tenant credit, duration and use case.

• Exaion majority-control deal closed in March; the team has begun contracts hosting U.S. inference and training workloads in Europe and is expected to support U.S. site development.

• Monetization paths include colocation and GPU-as-a-service; small sites may suit inference or neocloud deployments, while larger campuses can support multiple tenants and lease types.

Financing and project execution

• Starwood partnership is intended to reduce delivery risk; its track record includes over 7 GW of hyperscale data center capacity, and site contributions to joint ventures can limit incremental capital needs.

• Project-level funding was discussed as an 80% debt, 20% equity model; a theoretical 100 MW build was estimated at $1.2–$1.4 billion, or $10–$12 million per MW.

• Labor, transformers, switchgear, power access and permitting are key constraints; one panelist cited $12–$13 million per MW for projects turning on in 2027, with labor and equipment driving cost increases.