convex 🐸 🦙

My attempt to unpack what @mikealfred mentioned in the space last night on the significant opportunity for $IREN. IREN is best positioned IMO to server a new trend in Enterprise AI (think Fortune 500) customer needs: There is a trend within AI called “AI cloud repatriation” or “Infrastructure Sovereignty” where companies are moving their AI workloads OFF general public clouds, and bringing them “back” to private, dedicated infrastructure. Why? Because AI is becoming more and more ingrained into a company’s processes and workflows, to the point where the AI “harnesses” being built are the future fabric of the company. AI is literally becoming the core brain of the business. To put it simply: AI represents the company’s future IP, and there is a saying in enterprise IT “Don’t rent your brain” Therefore, many enterprise businesses (Fortune 500 plus heavily regulated industries and governments) will look to a fully vertical integrated AI data center solution for 3 reasons: 1) paranoia of shared hardware (multi-tendency): public clouds only separate companies at the software layer, but their data is still moving across the same physical network cables and storage drives. This is why companies like Jane Street and Pharma researchers are building their own AI DCs, their algorithm and research is their edge, so any risk greater than zero is unacceptable. 2) telemetry and metadata fear: even if a hyperscaler doesn’t train off the data directly, they still collect a lot of telemetry and metadata which is insightful: how many GPUs are you using, what times of day your inference spikes and in what countries / regions, etc. This is important when the Mag7 are also expanding into healthcare, finance, and logistics themselves 3) the threat of lock-in to one ecosystem: as mentioned before, AI is becoming the brain for companies to operate, so locking into 1 provider means you will be forced to pay higher and higher costs when prices go up, because the switching costs are too high when your company literally is running on your AI setup If this trend holds, IREN would be one of the FEW who could really capitalize on this, which the TAM would be the Fortune 500 companies plus heavily regulated / government / proprietary industries. It is just easier for a fully integrated DC provider to make these hardware specific adjustments to their DCs [networking, storage, etc], where speed and costs matter. One area of “weakness” for $IREN could be their software stack, but I believe solutions like Nvidia’s AI Enterprise solution will be able to help fill the void, if not direct M&A by IREN in the future. Therefore, don’t be disappointed if no “mega deal” is announced this week, but look at how they are strategically acquiring new power capabilities across the globe that will be key to their AI factory model to copy & paste the DC, but allow the tenant to customize / segment their data + networking + storage per their needs. I have been an $IREN holder since 2022, and am excited to see how this all plays out! #InIRENWeTrust Few

$IREN -> A few recent interesting quotes from Alex Freedland, the co-founder and CEO of Mirantis. This is a game changing acquisition for IREN. Full article below. "This isn't just about data privacy, though that's part of it," Freedland notes. "Countries, states, and industries are recognizing that AI is strategically important. They need sovereign approaches where the intelligence and the infrastructure producing it remain under their control." "Selling bare metal GPU (Graphics Processing Unit) infrastructure might generate $1 billion annually per 100 megawatts of power. Orchestrated, just-in-time compute services can generate $1.5 billion from the same capacity. But if you can package that as intelligence – actual AI services – you're looking at $4 billion from the same energy investment." "Energy constraints and sovereignty requirements mean the hyperscalers like AWS and Google won't have monopolies on intelligence production. Enterprises will consume AI services from multiple sources—hyperscalers for commodity intelligence, neoclouds for specialized capabilities, and potentially sovereign providers for strategic applications." markets.businessinsider.com/news/currencie…

$IREN Earnings Wish List for 5.7.26 (no particular order) After @danroberts0101 dropped two tweets within hours on Friday afternoon related to Sweetwater energization, the anticipation heightened going into the week leading up to Thursday earnings. Here's somethings on my mind that need more detail: 1) Revenue and EPS for QTR ending 3/31 Similar to @_Sgr_A_Star I expect a revenue and EPS miss due the bitcoin:native weakness in spot price as well as the removal of ASICs from Prince George, Mackenzie and possibly Childress that occurred by 3/31 quarter end. That said, the market likely anticipates this weakness already but will have its eyes glued to the AI quarterly revenue. "The Group reported $289,411,000 of unsatisfied remaining performance obligations, of which approximately $150,072,000 is expected to be recognized within the next 12 months." So the company had ARR of $150 million at year end and forecasted $500 million at the end of the quarter and reported $400 million contracted as of 2/5. So 500arr / 4 qtr = 125 qtr and 150/4 qtr = 37.5 (125+37.5) / 2 (midpoint) = 81.25m AI revenue estimated by GP The GPU delays that frustrated us with 17m the prior quarter may actually help us and result in additional revenue if $IREN secured better rental rates; that's the upside scenario and may boost revenue. 3) Prince George Confirm all of the GPUs or roughly 24k are installed and comment on if management secured stronger ARR than 500m for that batch. Also, any potential expansion of 30 MW at the site previously speculated by @FransBakker9812 4) Canal Flats Management removed from guidance in March and expect to mine bitcoin:native until September yet emphasizes time to compute. Is management waiting for Vera Rubin for this site? At roughly 15m/mw that a potential $400 million ARR if you assume 30 MW / 1.1 PUE at 15m/MW Long term its immaterial but for a company that just reported 17m AI revenue in the prior quarter, every converted MW matters and fast. 5) Mackenzie We know that the company increased guidance to $3.7b ARR in March with a 50k GPU purchase that $2.3B of the $3.5B purchase to Mackenzie or ~33k GPUs at $26,000 annually per GPU or $858 million ARR but that suggests an hourly rate of only less than $3.00. Management guided for a lower increase in pricing of 25% than the increased cost of B300 of 40-50%. Get aggressive Dan and up the guidance to reflect the current market's desperation for compute. That $858 million ARR becomes $1.14B with updated guidance to $4.00. Show your pricing power in this environment. Bonus points for announcing a deal and disclosing clients. 4) Childress - H1-4 Multiple items to address but the most important is updating the market on Horizon 1-4. Management will command a premium to its current valuation once Horizon 1 is delivered to $MSFT Ideally, we deliver H1 but at a minimum lay out the timeline of deliveries along timing on revenue to give the market more certainty. Bonus points for H3-4 substitution talks with $MSFT that show potential upside for ARR as well as access to Vera Rubin. 5) Childress - H5-6 Hash rate is dropping and rumors of mining halls being demolished. Is $MSFT or another hyperscaler asking $IREN to forego air cooled DCs with B300s for liquid cooled DCs that feature Vera Rubin? If so, announce the partner ideally and up guidance for pricing that justifies the opportunity cost of the lower retrofit cost of B300 6) Childress - 17k GPUs You allocated 17k GPUs for 26,000 per year or $442 million ARR. Same comments as Mackenzie; get more aggressive in guidance to reflect the market. $442 million likely becomes $550+ million ARR 7) The rest of Childress Currently, we know that there are 300 (H1-4) and 75 (17k GPUs) designated at Childress for almost a certain fact but that leaves 375 MW remaining. The company prides itself in optionality but that sometimes looks like indecision. A more defined path with additional ARR guidance will inspire more confidence in the site's vision. At 250 MW Critical IT (safely) that results in either the following a) B300s with lower cap ex for retrofit and faster delivery that likely command $10 million/MW annually or b) Vera Rubin delivered later for more cap ex but the upside of $15 million / MW with longer lasting infrastructure 8) Sweetwater We see the job postings and the 1.4 GW elephant requires commentary. Energization is awesome. How fast to we expect to get compute to market? What kind of client is right for the site or will there be multiple tenants? Bonus points for a deal. 9) Australia Lot of marketing spent down under with $MSFT and @AnthropicAI expansion to your hometown. Does $IREN have site(s) or will they pursue a joint venture? 10) More GPU purchases with more aggressive ARR guidance to reflect a market thirsty for compute that values time to compute more than ever 11) Oklahoma Simply confirm that things are on track for 2027-2028 and that your team is receiving interest all ready due to the compute constraints in the market 12) Pipeline I actually think it makes sense to keep this under wraps as the company focuses on the existing 4.5 GW with the exception of Australia due to the marketing dollars. This week is the reason we waited patiently. Enjoy it and let Dan and the team deliver. Best, BB

Sweetwater 1 has been successfully energized – a key milestone in the development of the broader 2GW Sweetwater campus. @danroberts0101, Co-Founder and Co-CEO of $IREN commented: “Delivering Sweetwater 1 substation energization on schedule reflects our disciplined execution, the strength of our supply chain relationships and the efficiency of our vertically integrated development model. It is another example of our ability to design and construct large-scale infrastructure reliably and at speed to meet market demand.” Learn more: iren.gcs-web.com/static-files/d…

If I had a direct line to @danroberts0101 (which I don't) I'd say this (after starting with thank you): i. You're solving for optionality at all times. ii. You have an open ATM. Even if you don't intend to draw on it, having a higher share price increases your optionality. iii. Time to compute is a great narrative but it's doing nothing for the stock. iv. Execution has been and will continue to be your greatest differentiation point. Use it. v. Don't wait for May earnings - as you hit milestones throughout Q2 (and beyond), blitz the market with PRs. Why? You will differentiate on proven, consistent execution that competitors cannot match. vi. Horizon 1 delivered? PR. vii. Sweetwater 1 energized? PR. viii. PG and/or Mackenzie (or even Canal) fully contracted? PR. ix. First contract signed for air cooled AI Cloud capacity in Childress? PR. x. Australian datacenter expansion? PR. xi. Datacenter financing for Horizons and/or incremental GPU financing or corporate level debt secured? PR. xii. Sweetwater deal? PR. xiii. Undisclosed sites included in Batch Zero? PR. xiv. Execution in a vacuum lets FUD breathe and reduces optionality. I get that you're hitting a scale where certain developments (e.g., the 10 MW site at PG) are no longer material to the business as a whole, but when the ATM is open, your optionality increases as momentum on execution is made clear. IR is a weapon, not a distraction.