IREN Limited mines Bitcoin and sells AI computing power from the same renewable-energy grid connections, shifting megawatts between banks of ASIC miners and NVIDIA GPU clusters depending on which is more profitable at any given moment. Each site has a hard power ceiling set by its grid-connection agreement, so every watt running a GPU is a watt not mining Bitcoin — and that zero-sum tension is the operating lever the whole business turns on. Because the site infrastructure is already permitted, cooled, and wired for both types of hardware simultaneously, a new competitor cannot replicate the setup just by buying equipment — the permitting and dual-load buildout has to come first, which capital alone cannot accelerate. The same constraint that creates the advantage also creates the fragility: if a curtailment order or policy change cuts the available megawatts at a site, there is no longer enough power to run both workloads at meaningful scale, and the reallocation mechanism that separates IREN from two ordinary single-purpose data centers disappears with it.
How does this company make money?
When the company's ASIC miners run, they earn Bitcoin rewards based on how much of the total Bitcoin network's computing power the company contributes. When the GPU clusters run, customers pay for compute time by the hour, or pay colocation fees to house their own hardware at the site. The company earns more from whichever workload is paying better at a given moment, and the whole business is built around shifting power between the two to chase that higher return.
What makes this company hard to replace?
AI customers running workloads on NVIDIA reference architecture would have to go through hardware requalification cycles before their software could run reliably on a different provider's setup. Bitcoin mining operations cannot simply move to a new location without losing accumulated network uptime, which affects how the operation is treated by the Bitcoin network. The physical location of renewable energy sites also creates geographic constraints — not every alternative facility sits on the same grid resources.
What limits this company?
Every site has a maximum megawatt ceiling set by its grid-connection agreement. No amount of money can push past that ceiling without a new interconnection agreement, and getting one requires site-specific permitting that takes time no matter how much capital is available. Until a new agreement is in place, adding more ASIC miners or more NVIDIA GPU racks is physically impossible — there is simply no power left to run them.
What does this company depend on?
The company cannot run without grid-connected renewable energy sites that provide its power ceiling, ASIC mining hardware that does the Bitcoin work, NVIDIA GPU hardware and reference architecture for AI compute, fiber optic connectivity to move data in and out of the sites, and the Bitcoin network protocol that determines mining rewards.
Who depends on this company?
The Bitcoin network loses hash rate — meaning it becomes slightly less secure — if the mining operations stop. AI training customers lose access to their GPU compute clusters and would need to find alternative infrastructure. High-performance computing workloads that run on the GPU clusters would go dark until replacement infrastructure could be found.
How does this company scale?
Within a site that already has its grid connection, adding more mining hardware or more GPU racks scales in a straightforward way — buy the hardware, install it, run it. But growing beyond existing sites requires finding new grid-connected renewable energy locations, going through site-specific permitting, and building dual-load infrastructure from scratch. That process cannot be made faster by spending more money, which means the number of viable sites the company can operate is the real growth ceiling.
What external forces can significantly affect this company?
The Bitcoin network automatically adjusts how hard it is to mine Bitcoin based on how many miners are competing, which changes how much revenue each megawatt of mining actually produces. Renewable energy policy in any jurisdiction where the company operates can change or revoke the grid access the entire business depends on. When NVIDIA faces chip shortages, it sets its own priorities for who gets GPU allocations, which can delay or block the company from expanding its AI compute capacity.
Where is this company structurally vulnerable?
If a government in one of the company's operating jurisdictions issues a curtailment order or cancels the interconnection agreement that sets a site's power ceiling, the total megawatts available drop below what is needed to run both ASIC miners and NVIDIA GPU clusters at meaningful scale. At that point the reallocation mechanism — the only thing that makes these sites different from two ordinary single-purpose data centers — stops working because there is nothing left to reallocate.