Paul Golding

Les walked through AMD's two options (50 MW expansion to 100 MW, plus a further 100 MW subject to guardrails) and declined to speculate on how the new lease affects AMD's decisions.

Management clarified the expansion timeline is broadly in line with the initial phase, noting prior phases had early preparation before announcement. ROFR-to-option conversion simplifies expansion while clarifying discussions with other potential Rockdale tenants.

Management explained same rental rate with escalator clause driving higher TCV, combined with lower build cost from leveraging building preparation already completed in the initial phase.

Jason Chung described a two-tranche financing approach: low-cost facility for stabilized 25 MW and separate delayed-draw for 75 MW expansion, avoiding blended construction premium and enabling equity recycling.

Management cited the AMD delivery as validating both their commercial approach and the strategic advantage of internal engineering capabilities through ESS Metron, describing it as a replicable playbook for future leases.

Management framed Core & Shell as derisking the project for hyperscalers by demonstrating approved power, experienced team, balance sheet support, and construction already underway.

Management noted ESS Metron serves major external data center projects nationally with Riot representing a small portion, and the growing backlog reflects ability to balance both internal and external demand.

Management reaffirmed build-to-suit as value-maximizing but clarified they will not build on spec without a lease, having already invested in foundational infrastructure.

Management confirmed water line plans approved for full 1 GW development, retention pond providing natural water, and critical substation equipment already arriving to derisk timeline.

Management clarified 600 MW is primarily available but the full 1 GW could be on the table if economics support conversion.

Management cited ESS Metron as a key differentiator for supply chain control, with positive feedback from counterparties on this capability.

Harry said meeting even bear/mid-case data-center growth will require behind-the-meter deployments, which CleanSpark is actively evaluating.

Harry said the build is a true greenfield with only modest accretive value from the already-built, energized substation.

Harry Sudock described how early diligence and utility relationships in the Greater Houston market let CleanSpark close the first 300 megawatts at full approval with the second 300 deeply progressed; a thorough, direct answer.

CEO said it is early to comment on deposit/prepayment terms but signaled it aims to set a better lease standard than the market expects; partial answer.

Gary said the Bitcoin treasury strategy is unchanged: monetize via yield strategies and borrow against it for accretive acquisitions.

Golding asked about hash-rate pricing as peers pivot to HPC/AI; Bradford said softening miner demand creates countercyclical buying opportunities.

Zach said the prepaid infrastructure is largely immersion cooling, which they view as the best long-term path; direct.

Kent Draper said IREN is doing both, building data centers and buying compute on spec and against contracts.

Kent Draper said the pivot is demand-driven, with an effective liquid-cooling design for BC conditions.

Kent Draper said air-cooled Blackwells perform well; on the follow-up he noted air-cooled operating margin is slightly higher than liquid-cooled but immaterial, with the real benefit being capital efficiency of retrofitting existing sites.

Kent cited jurisdictional diversity, low-cost renewable power, and hyperscaler interest in the region.

Demand is dominated by bare-metal customers; software is a smaller, likely-commoditized layer they monitor.

Dan confirmed strong, rising demand but declined to specify dollars per GPU hour, redirecting to the risk-return proposition and levered IRRs.

Kent Draper gave specific PUE figures (1.1 air-cooled BC, under 1.2 with liquid cooling, ~1.2 average at Horizon 1) and said deployments are customer-led, not PUE-led.

Any deal can fall within the Starwood partnership; only the portfolio at signing was pre-allocated.

Deal has contingencies tied to batch-zero approval with a terminal close-or-walk point, structured to benefit MARA and align counterparty interests.

Management cited broad demand across multiple sites and expects to progress through the ERCOT queue as the audit process advances.

Fred said Long Ridge had been in the works since the Hannibal acquisition and framed a dual approach: small tuck-in inference sites plus larger campuses developed with Starwood.

Fred described a spectrum from hyperscaler training/inference to enterprise token-factory and private-cloud needs, expecting inference volumes to rise sharply with agentic adoption.

Fred explained MARA contributes the site as its initial asset plus capital for its share of development costs, retaining up to 50%; deferred some mechanics to later.

Michael said stablecoin and crypto payments are nascent pending regulatory clarity, but Mastercard has a natural role in interoperability, safety standards, and identity. Economics still to be determined.

Ryan said AI both drives internal productivity and differentiates VAS products (Featurespace, data token).