I’ve been analyzing IREN’s expansion and financial metrics, and I’d love to get your insights on the following questions:
1️⃣ Mining Revenue & Costs
If IREN achieves 50 EH/s with 810 MW (660 MW + 150 MW expansion) and the network hash rate is 1,000 EH/s, they would mine approximately 8,350 BTC per year.
Given 810 MW and 8,350 BTC per year, does the estimated yield of 10.48 BTC per MW per year seem reasonable?
If Bitcoin price = $100K, then 1 MW generates approximately $1M in revenue per year.
3️⃣ AI/HPC Data Center Classification
IREN’s 75MW AI/HPC data center is expected to cost $300M to build.
Given an estimated $4M per MW, would this classify as a Tier 2 (T2) data center?
Typically, Tier 3 (T3) facilities have higher redundancy & uptime guarantees, increasing costs. Does this align with industry pricing for Tier 2 facilities?
4️⃣ Expected AI Hosting Revenue Per MW
If IREN’s AI data center is Tier 2, expected revenue per MW should be $1.5M annually, based on comparisons to Core Scientific (CORZ), which earns $1.44M per MW from CoreWeave.
Does $1.5M per MW seem achievable in IREN’s AI colocation model or would it be higher?
5️⃣ Operating Costs Per MW for AI Data Center
what would be Estimated operating costs per MW per year:
Electricity: $30K per MW?
Cooling: $$50K per MW?
Maintenance: $15K–$25K per MW
Network & Connectivity: $10K–$30K per MW
Depreciation (Infrastructure): ?
6. How does IREN compare with CORZ in terms of AI//HPC. CORZ will have ARR of $360M (250 MW)starting from 2026 and $850M (590 MW) starting from 2027.
1-2. Iren gives a guidance on this specifically on their monthly reports (with sensitivities on BTC price) that I found very useful for this. The baseline scenario they have is $977M revenue @ $105k BTC, that would imply 9,300 BTC per year. That will be with the 810 MW.
You will then have halving and all the stuff with time, you know? But BTC price might change too, so I think you just take that at face value and may be play around a bit with a lower BTC price in case you want to be conservative.
3. Guidance has been $300-350M and when you look at $ per IT Load (50 MW), that's around $6 to $7m CAPEX per MW. This is much higher CAPEX vs. other developments at Childress. I'm pretty sure Horizon 1 is being built with Tier 3 characteristics, Kent (CCO) has confirmed this in interviews too.
And I am no industry expert by any means lol, but I asked Gemini and the cost for a Tier 3 is from $7 to $12 million per MW (all in). But then, take into account that Iren is investing $6-7 MW on a land that has already the power access. Valuing access to power is more art than science, but when you factor that the all in cost seems roughly comparable to a Tier 3, don't you think?
4- I believe revenues from Horizon 1 will be around $60-100M monthly, which would be $1.2-2.0M per MW (IT Load). I did the estimates based on the deal anounced by Wulf, you can check it out here: https://carlosfsaavedra.substack.com/p/iren-investment-thesis. If you do the math with CORZ numbers, they are pretty similar. So yeah, $1.5M per MW seems pretty reasonable to me.
5- I haven't done my homework on this one!
6- I think being hybrid is the actual advantage you have with Iren. At current context, BTC offers roughly $1.3M revenues per MW with >70% profit margins, though you do need to factor depreciation in the equation because mining equipment needs to be replaced pretty fast. The main thing is that, either because Iren is the only profitable all-in cost miner or because of the industry itself, the market does not price this at all. So, multiples are compressed and you can't finance mining equipment. But that's pretty different with AI / HPC. In this case, you find a customer, sign a contract for 5, 10, 20 years and, even if the revenues $ per MW are similar / just slightly higher, the doors of financing open up for you. The contracts are fixed and it will ultimately depend on the credit of the customer, but that's a plain vanilla financing in the real estate world. You got a customer with good credit that signs 5-10 year lease that is fixed with no unilateral early termination, and that's something you can 100% finance. That's an Equinix REIT-like model that values your EBITDA at 20x since it is recurring and predictable.
But you still got a company that, at current conditions and excluding Horizon 1, will do roughly $1B annual revenues, $600M EBITDA, and operate profitably at a net profit margin with equipment depreciation. So there is no hurry on making the AI / HPC transition because the opportunity cost is there. Still, when you hear Dan (co-CEO) in the interviews, it seems the demand is so high that is likely that Childress (where BTC mining is) will be retrofitted to support new "Horizons" prior to Sweetwater 1 getting live in April 2026.
So end of the day, when you look at Iren AI/HPC right now, you don't have those secured revenues CORZ or other players have. But the infrastructure is there and the management have given strong guidance on positive negotiations. And that's why you got the assymetric opportunity with Iren right now you don't have with other companies (including CORZ) in my opinion.
1. at 909 network hash rate, its around 9300 bircoins and when with 1000 hash rate, it will be 8350 bitcoins.
2. do you think by next year, they will have colocation deal for 1.4 GW sweetwater 1 DC. i think that is main catalyst and with AI/HPC revenue will be around 2 Billion?
I’ve been analyzing IREN’s expansion and financial metrics, and I’d love to get your insights on the following questions:
1️⃣ Mining Revenue & Costs
If IREN achieves 50 EH/s with 810 MW (660 MW + 150 MW expansion) and the network hash rate is 1,000 EH/s, they would mine approximately 8,350 BTC per year.
Estimated annual costs: $ 355M (Electricity: $235M Overhead + Energy Certificates: $120M)
2️⃣ Bitcoin Yield Per MW
Given 810 MW and 8,350 BTC per year, does the estimated yield of 10.48 BTC per MW per year seem reasonable?
If Bitcoin price = $100K, then 1 MW generates approximately $1M in revenue per year.
3️⃣ AI/HPC Data Center Classification
IREN’s 75MW AI/HPC data center is expected to cost $300M to build.
Given an estimated $4M per MW, would this classify as a Tier 2 (T2) data center?
Typically, Tier 3 (T3) facilities have higher redundancy & uptime guarantees, increasing costs. Does this align with industry pricing for Tier 2 facilities?
4️⃣ Expected AI Hosting Revenue Per MW
If IREN’s AI data center is Tier 2, expected revenue per MW should be $1.5M annually, based on comparisons to Core Scientific (CORZ), which earns $1.44M per MW from CoreWeave.
Does $1.5M per MW seem achievable in IREN’s AI colocation model or would it be higher?
5️⃣ Operating Costs Per MW for AI Data Center
what would be Estimated operating costs per MW per year:
Electricity: $30K per MW?
Cooling: $$50K per MW?
Maintenance: $15K–$25K per MW
Network & Connectivity: $10K–$30K per MW
Depreciation (Infrastructure): ?
6. How does IREN compare with CORZ in terms of AI//HPC. CORZ will have ARR of $360M (250 MW)starting from 2026 and $850M (590 MW) starting from 2027.
Hey Anki! Thanks for your comment.
1-2. Iren gives a guidance on this specifically on their monthly reports (with sensitivities on BTC price) that I found very useful for this. The baseline scenario they have is $977M revenue @ $105k BTC, that would imply 9,300 BTC per year. That will be with the 810 MW.
You will then have halving and all the stuff with time, you know? But BTC price might change too, so I think you just take that at face value and may be play around a bit with a lower BTC price in case you want to be conservative.
3. Guidance has been $300-350M and when you look at $ per IT Load (50 MW), that's around $6 to $7m CAPEX per MW. This is much higher CAPEX vs. other developments at Childress. I'm pretty sure Horizon 1 is being built with Tier 3 characteristics, Kent (CCO) has confirmed this in interviews too.
And I am no industry expert by any means lol, but I asked Gemini and the cost for a Tier 3 is from $7 to $12 million per MW (all in). But then, take into account that Iren is investing $6-7 MW on a land that has already the power access. Valuing access to power is more art than science, but when you factor that the all in cost seems roughly comparable to a Tier 3, don't you think?
4- I believe revenues from Horizon 1 will be around $60-100M monthly, which would be $1.2-2.0M per MW (IT Load). I did the estimates based on the deal anounced by Wulf, you can check it out here: https://carlosfsaavedra.substack.com/p/iren-investment-thesis. If you do the math with CORZ numbers, they are pretty similar. So yeah, $1.5M per MW seems pretty reasonable to me.
5- I haven't done my homework on this one!
6- I think being hybrid is the actual advantage you have with Iren. At current context, BTC offers roughly $1.3M revenues per MW with >70% profit margins, though you do need to factor depreciation in the equation because mining equipment needs to be replaced pretty fast. The main thing is that, either because Iren is the only profitable all-in cost miner or because of the industry itself, the market does not price this at all. So, multiples are compressed and you can't finance mining equipment. But that's pretty different with AI / HPC. In this case, you find a customer, sign a contract for 5, 10, 20 years and, even if the revenues $ per MW are similar / just slightly higher, the doors of financing open up for you. The contracts are fixed and it will ultimately depend on the credit of the customer, but that's a plain vanilla financing in the real estate world. You got a customer with good credit that signs 5-10 year lease that is fixed with no unilateral early termination, and that's something you can 100% finance. That's an Equinix REIT-like model that values your EBITDA at 20x since it is recurring and predictable.
But you still got a company that, at current conditions and excluding Horizon 1, will do roughly $1B annual revenues, $600M EBITDA, and operate profitably at a net profit margin with equipment depreciation. So there is no hurry on making the AI / HPC transition because the opportunity cost is there. Still, when you hear Dan (co-CEO) in the interviews, it seems the demand is so high that is likely that Childress (where BTC mining is) will be retrofitted to support new "Horizons" prior to Sweetwater 1 getting live in April 2026.
So end of the day, when you look at Iren AI/HPC right now, you don't have those secured revenues CORZ or other players have. But the infrastructure is there and the management have given strong guidance on positive negotiations. And that's why you got the assymetric opportunity with Iren right now you don't have with other companies (including CORZ) in my opinion.
Thanks Carlos for all the insights.
1. at 909 network hash rate, its around 9300 bircoins and when with 1000 hash rate, it will be 8350 bitcoins.
2. do you think by next year, they will have colocation deal for 1.4 GW sweetwater 1 DC. i think that is main catalyst and with AI/HPC revenue will be around 2 Billion?