Steve Glagola

Salman framed the ATM as one of several opportunistic, KPI-driven capital sources, stressing a high return-on-capital-employed threshold and accretive deployment rather than a direct ATM-versus-leverage comparison.

Fred said 2PIC immersion lets MARA overclock miners meaningfully, cutting rig count and CapEx by roughly a third, and is being adapted as liquid cold-plate cooling for AI GPUs via compute-OEM pilots.

Salman said hash cost should keep declining as low-cost owned generation like the 114-megawatt wind farm comes online, prioritizing stockholder value over single KPIs.

Fred said MARA is executing on energy-company partnerships across the U.S., Europe and the Middle East using a mix of thermal, wind, solar and flare gas, but declined to give specifics.

Fred said the market credits MARA's HODL but not its mining operations, and argued state and institutional Bitcoin adoption is reshaping the price cycle.

Fred reaffirmed the full HODL strategy adopted in July 2024, describing a twin-turbo model that captures both low-cost mining and Bitcoin price upside on over 48,000 coins.

Gary targeted above 90% loan to cost, cited Alphabet's oversubscribed bond deal as evidence of robust demand, and noted hyperscaler spreads tightening.

Matt candidly said there is little certainty on what follows the August 20 PUCT meeting, but the governor is signaling data centers are a Texas priority.

Management said the ERCOT process is not final but detailed why its Texas sites are well-positioned (completed studies, executed interconnects); thorough answer.

Glagola asked whether any sites suit AI/HPC given peers' higher multiples; Bradford said mining's consistent 55% returns set a high bar for any alternative use.

Zach and Gary said each added exahash already produces free cash flow to the bottom line, that capital return via buybacks is possible once the balance sheet is fortified, and that timing will be driven by input costs.

Kent said 11,000 of 23,000 GPUs were contracted plus ~1,000 more since; uncontracted units are later-delivery, with strong demand and late-stage discussions on the rest.

Kent Draper said IREN is not yet on Lepton but evaluating it as an additional demand driver reaching smaller AI start-ups/enterprises; did not disclose fee/take-rate economics the analyst asked about.

Dan dismissed both concerns, arguing West Texas AI suitability is settled by industry announcements and that IREN has run NVIDIA GPUs beside miners for 12 months, addressing the miner-credibility worry.

Management confirmed leadership transitions were planned enhancements with no disruption, citing AMD's option exercise as evidence of continued execution, and affirmed the team is already executing at both sites concurrently.

Jason Chung described funding hierarchy starting with Bitcoin treasury sales plus monthly production, supplemented by low-cost non-dilutive debt structures as data center development opens new capital pools.

Management noted existing FEAs insulate both sites from renegotiation, with potential 4CP program changes being monitored but too early to assess impact on transmission charges.