[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]  Dulce [@litigious_dulce](/creator/twitter/litigious_dulce) on x 2639 followers Created: 2025-07-14 05:10:46 UTC I was just relying on memory at the time and may have been wrong. Let’s assume X year contract for the B200s (pricing varies depending on contract duration, with on demand being the most expensive and X years being the cheapest). Lambda charges $XXXX per gpu-hr for a X year contract. Lambda is representative of the market, so let’s just say $IREN charges $XXXX per gpu-hr, all costs included (many CSPs charge costs in addition to gpu-hr). If IREN is using a CSP aggregator like Nvidia Lepton (“the Uber of GPU Capacity”), then we can assume virtually XXX% utilization. So $3.25*24*365=$28.5k. If IREN spent $40k per B200 (including cost of retrofit, infiniband, etc.), then the payback period is XX months… IREN will have about XXX MW of data centers in Canada. If IREN uses all XXX MW for CSP, then IREN can deploy about 100k B200s. The ARR would therefore be $2.85B. Should be fairly easy to calculate EBIT or net profit of CSP from here. ARR obviously depends on $/gpu-hr. To have $4B ARR, you need on average $4.5/gpu-hr, which may be doable depending on the mix of on-demand and contracted GPUs. For example, Nebius charges $5.5/ gpu-hr for on demand. I described CSP aggregators as having no customer acquisition cost, but that’s not exactly true. Conceptually, I think the CAC is the margin IREN yields to the CSP aggregator. However, IREN’s CSP is not mature or reputable enough where it can demand a premium. Therefore, maybe in theory $IREN could make $3.75/gpu-hr if IREN directly negotiated with clients. However, IREN is likely ok with giving up $0.50-0.75/gpu-hr in the short-term to guarantee immediate and scalable utilization of its GPUs. Over time, IREN will likely develop its software stack. Then, once IREN is perceived as a top-tier CSP, IREN’s CAC will be low, such that a middleman/broker will no longer be necessary. IREN can form direct relationships with clients and increase profit margins. That said, if software is commoditized as we predict, then even in that scenario CSP pays very well. XXX engagements  **Related Topics** [$iren](/topic/$iren) [Post Link](https://x.com/litigious_dulce/status/1944625596421767229)
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
Dulce @litigious_dulce on x 2639 followers
Created: 2025-07-14 05:10:46 UTC
I was just relying on memory at the time and may have been wrong.
Let’s assume X year contract for the B200s (pricing varies depending on contract duration, with on demand being the most expensive and X years being the cheapest). Lambda charges $XXXX per gpu-hr for a X year contract. Lambda is representative of the market, so let’s just say $IREN charges $XXXX per gpu-hr, all costs included (many CSPs charge costs in addition to gpu-hr).
If IREN is using a CSP aggregator like Nvidia Lepton (“the Uber of GPU Capacity”), then we can assume virtually XXX% utilization. So $3.2524365=$28.5k. If IREN spent $40k per B200 (including cost of retrofit, infiniband, etc.), then the payback period is XX months…
IREN will have about XXX MW of data centers in Canada. If IREN uses all XXX MW for CSP, then IREN can deploy about 100k B200s. The ARR would therefore be $2.85B. Should be fairly easy to calculate EBIT or net profit of CSP from here.
ARR obviously depends on $/gpu-hr. To have $4B ARR, you need on average $4.5/gpu-hr, which may be doable depending on the mix of on-demand and contracted GPUs. For example, Nebius charges $5.5/ gpu-hr for on demand.
I described CSP aggregators as having no customer acquisition cost, but that’s not exactly true. Conceptually, I think the CAC is the margin IREN yields to the CSP aggregator. However, IREN’s CSP is not mature or reputable enough where it can demand a premium. Therefore, maybe in theory $IREN could make $3.75/gpu-hr if IREN directly negotiated with clients. However, IREN is likely ok with giving up $0.50-0.75/gpu-hr in the short-term to guarantee immediate and scalable utilization of its GPUs.
Over time, IREN will likely develop its software stack. Then, once IREN is perceived as a top-tier CSP, IREN’s CAC will be low, such that a middleman/broker will no longer be necessary. IREN can form direct relationships with clients and increase profit margins. That said, if software is commoditized as we predict, then even in that scenario CSP pays very well.
XXX engagements
Related Topics $iren