Mark Laird

$IREN VR200 boosts FWD ARR from $3.7B to $5.8B WARNING: Speculative Post NFA DYOR Recent monitoring of $IREN hash rate implies more bitcoin:native ASICs are coming offline and likely will be replaced by GPUs at Childress. Much of my analysis lately focused on the potential of air cooled B300s that require lower retrofit costs and speed up time to compute as $IREN scales ARR, but let's not forget @danroberts0101 loves himself some optionality. Is it possible, for the right price, that Dan would agree with $MSFT to upgrade GB300s to VR200 NVL 72 (the newest bad ass $NVDA GPU that increases output 3.5x and 5x per rack in training and inference respectively)? I linked a post by @FransBakker9812 below describing the ramifications but here's the TLDR; $IREN and $MSFT can upgrade Horizon 3 and 4 from GB300 to VR200 NVL 72 if both parties can mutually agree upon revised terms. Management prudently protected itself from a forced upgrade by $MSFT as both parties need to agree updated pricing in consideration of the higher GPU cost of VR 200 of approximately $122,000 based on a conservative estimate from tomshardware.com $IREN built flexible racks with Rubin always on their mind that could flex from 130 to 200 kw rack densities. Said another way, $IREN can fit the same number of Rubins as GB300s despite the higher power draw due to the flex design and purposely used fewer GB300s. Knowing all that; check out my analysis below and here are the highlights: 1) Breakeven Scenario - $iren charges $3.95/hr to recoup the cap ex increases from substitution 2) Constant Margin - Assuming the same Gross Margin % of 40.2% the hourly rate increases to $4.67/hr. I find this case the most likely and it would increase ARR by $595m for H3-4 and $1.566B if $iren gets $msft to agree to H5-6 (part of the initial plan before air cooled discussions) 3) Margin pickup - This scenario increases gross margin % to 44.2% and has potential to come to fruition as execution risk is gone after imminent H1 completion and compute remains hard to find, especially liquid cooled according to @kentpdraper. Finally, consider $IREN current guidance of $3.7B ARR, the incremental revenue from substitution at 3-4 as well as the Horizon 5-6 expansion at $4.67 per hour would result in increases of $595m and $1.566B respectively. 3.7B+0.595B+1.566B=$5.861B Assuming $IREN can secure supply from $NVDA than argument for moving forward with VR 200 and liquid cooled DCs can be justified assuming higher prices but also an understanding that the new data center life will extend at least 15-20 years beyond the initial contract and avoid a future upgrade cycle following the obsolescence of the B300. h/t @_Sgr_A_Star @Agrippa_Inv for contributions

H/t @callah32 What bubble? $iren $nbis $crwv