Mackenzie UPDATE: $IREN looks to be laying the groundwork for liquid cooling at Mackenzie while the air-cooled retrofit is still underway. Drone footage from 9/23 shows black 2" to 3" rigid pipe entering the data hall wall at regular intervals along an open trench that runs the length of the building. It looks like supply and return branch lines for a liquid cooling system. If you look closely, they go from the wall into buried trunk lines running parallel to the data hall for a future liquid-cooling loop, going in at the same time as the air-cooled GPU fit-out. Why this is smart sequencing:

IREN is partnering with NVIDIA to launch Minerva, NVIDIA’s new AIOps platform, on IREN AI Cloud. Our VP of AI Cloud, Jason Forrester, joins NVIDIA’s Gilad Shainer and 650 Group’s Alan Weckel to explore how IREN’s approach to AIOps enhances visibility, strengthens operations and keeps compute productive. Watch the conversation: youtu.be/MWVjiWYs1sw

IREN’s @danroberts0101 speaks to @ArjunKharpal about AI financing and demand on CNBC. Watch the full interview.

A lot of questions this week on funding, so here's how a GPU typically gets funded at $IREN. First, the customer signs a multi-year take-or-pay contract and agrees to a prepayment. On our recent 3-year contracts, prepayments have been 45% to 55% of the associated GPU capex. Second, financing is committed, secured by the contracted GPUs. On the Microsoft contract we secured an A-rated facility at around 6%, which together with prepayments covers 96% of GPU capex. More recently, for a non-investment grade counterparty at Mackenzie, we secured financing at 9% fixed, which together with prepayments covers more than the full cost of the GPUs. That's the GPU side. The data centers are a separate, long-term asset base, and we own the ones our GPUs run in. Construction has been funded from cash to date, with no debt against these multi-billion dollar assets. As they come online with contracted customers, they can support financing in their own right. The processes behind the next facilities are continuing, and new paper will cost what the market charges. With GPU paybacks of around two years on our recent 3-year contracts, a wider spread barely moves the economics. There's more customer demand than we can build for. As we noted at our recent earnings, 2026 is largely sold out, recent 3-year contracts are at more than double the prior year's pricing, and demand for anything we can bring online in 2027 covers it several times over, with prepayments a focus in every negotiation. Conversations on 2028 capacity are already underway. Meanwhile announced capacity keeps slipping across the industry, which is the point: signing deals isn't the bottleneck in this market, bringing GPUs online is, and that's our focus.

$IREN Status of H2 Timelines From two separate sources, one is a 2nd degree source and the other is a 3rd degree source, MSFT is requesting changes to H1's networking topology. By extension, this topology will be need to be cabled for H2-4. The re-cabling on H2 was given a timeline of 45 days while H1 will be re-cabled while running workloads aka billing and that process was given a timeline of 4 months. H3 will be delivered with H4 as a supercluster, H4 hasn't begun cabling and will maintain December timeline. H1 recabling will take much longer than H2 because H1 is already accepted so the changes will be made during maintenance window in stages. H1 will be running live workloads and billing. I'm not sure when the 45 days started but H2 will be delivered either late Nov or Early Dec. Both H3 and H4 will be delivered as a super cluster Dec. What's the Issue This request comes after acceptance test while running workloads on H1 but before the supercluster networking before H1 and H2. The issue was described as networking bandwidth between racks within a scaleable unit. All of H1 counts 1 scalable unit. Networking physical distances are not as sensitive as Cray supercomputer days and spec is around 50 meters. This is important because it means that racks will not need to be moved, aka no changes to liquid cooling pipes or power distribution components. Cabling does take a long time because there are literally hundreds of thousands of cables that need to carefully cabled as wrong cable will take more time debugging than getting it right in the first place. The cabling team used for H1-4 too. However, internally the deadline is still 2026. Who is at Fault MSFT is a very large company with multiple team. Their bare metal team signed off because IREN met Nvidia DSX specs. However MSFT platform team when deploying software need a specific topology for their workload and was not communicated to the bare metal team. This explanation is based on the fact that this was not a test case in acceptance testing and found more than 1 month after acceptance and usage began. During this one month, it's likely that platform team deployed and only now did MSFT's AI research team started running workloads. Between IREN, Dell, MSFT - it's most likely MSFT's platform team did not gather requirements from MSFT's research team and did not relay the full requirements to MSFT's bare metal team. Dell is paid for cabling based on MSFT specs. And no, there cannot be one huge stress test to have maximized traffic between every rack within the SU because the networking bandwidth is maximized for workload traffic patterns. Every rack pair cannot simultaneously get each rack’s full bandwidth. Different topology maximizes for different workloads. IREN's supercluster topology is most likely rail-optimized, non blocking fat-tree.

Know what you own. Flushes like this are not by choice. Decade long data centers are not weekly investments. Stay strong frens. $IREN