I talked about Neoclouds extensively at the end of 2025, including deep dives into CoreWeave and Nebius. Now it's time to veer into IREN right before its Q2 earnings today. [Fingers crossed it's a good one.]
IREN, at its core, is a company focused on data center design & optimization. Thus far, they used those talents to become one of the largest and most efficient Bitcoin miners. It has now begun to leverage those core talents into a new direction. After years of building up expertise in building data centers near cheap, reliable energy sources, it is now pivoting that wealth of experience into becoming a neocloud offering bare-metal GPUaaS.
Q226 earnings are tonight after the close. [Unfortunately, my write-up and editing work took me right to the wire on Q226 earnings.]
It seems they pushed their earnings date up a week this year from past ones (~Feb 12-15 to now Feb 5), sparking FinTwit speculation of an imminent new whale deal coming for Sweetwater (due to energize in Apr-26, and start unfolding in phases from there). Maybe the answer is simpler – they might be establishing a new earnings cadence, given they have a brand new CFO (promoted their new Chief Capital Officer after a few months), plus are still new to US GAAP reporting (started July 1).
The "imminent" part is assuredly true given the capacity rollouts about to unfold over the coming year, but if the next deal announcement is not today, it may cause a further negative reaction until it finally appears. This is exacerbated by the drop in BTC price over Q4 2025 (showing up in today's Q226 earnings) that looks to be much worse in Q1 2026 (showing up in their next Q326 earinings). AI revenue is rapidly increasing over the coming quarters, but will not be enough to offset the revenue loss in Q3 should BTC level out or worsen.
Sections:
- IREN's neocloud ambitions
- Different approaches between CoreWeave, Nebius, and IREN
- The rise of their mining
- The phased approach to Childress
- The pivot to Horizon
- The Microsoft whale deal
- British Columbia retrofit
- The coming Sweetwater Hub
- GPU growth trajectory
- Funding sources
- Levers from here
This will be of interest to other neoclouds like CoreWeave and Nebius, the hyperscalers, and AI chipmakers like NVIDIA and AMD.
- IREN is one of the rare cryptominer-to-AI plays that interest me, as it wants to become a neocloud, not a data center landlord to hyperscalers and neoclouds. And while they derive revenue from crypto mining, they don't carry the resulting mining rewards (BTC) on their balance sheet.
- They have visible DC buildout & operational experience, with an existing footprint of 6 DC sites with 810MW active capacity and ~3GW of potential capacity through existing grid connections. This gives them a big head start in the land and power needs that every hyperscaler, neocloud, and cryptominer are battling over.
- In their massive cryptomining expansion across 2024-2025, mgmt built an operational machine capable of building a 50MW data center a month (+150MW a quarter) as they expanded Childress from its 20MW start to full 750MW potential. I believe they will repeat this cadence with Horizon and Sweetwater buildouts, with a pathway to phase in large blocks of GPUs over the next few years without any new sites.
- IREN hits its buildout targets. I go into extensive detail below on how the phased buildout of Childress was guided and executed.
- After a successful year+ of massively scaling up their mining capacity and efficiency, they paused cryptomining expansion in Mar-25, and are now all-in on neocloud from here.
- They are now focused on 2 new massive ongoing buildouts (Horizon and Sweetwater), plus are also retrofitting their 3 BC sites (160MW) – with displaced mining rigs sent to Childress.
- Their history thus far is in cryptomining for themselves (no customers). They have shown they can execute on their buildouts & retrofits, so the real challenge from here is in selling it to hyperscalers or frontier AI firms under long-term contracts and structuring deals to maximize profitability.
- They recently landed a major contract with Microsoft, who is now renting bare metal capacity from them, in addition to its ongoing deals with neocloud competitors CoreWeave and Nebius. Microsoft did its diligence and believes IREN is up to the task, noting their buildout speed as a factor.
- But most of all, this Microsoft deal gives IREN a much-needed boost in brand perception, helping counter the negative "Underperforming" rating that IREN got in ClusterMAX 2.0 in November. (That rating seems mostly due to the lack of software stacks atop the cloud GPUs.)
- They are about to see a massive ramp-up in AI revenue. Mgmt is signalling AI revenue will hit a $200-$250M annualized run rate at the end of 2025 (their Q226, aka today's earnings), $500M in early 2026 (their Q326), and $3.4B by the end of 2026 (their Q227). This implies AI revenue will hit $50-62M in Q226, $125M in Q326, and $850M in Q227.
- Unfortunately, this is more than offset by some significantly poor market dynamics in the Bitcoin market. They will see a drop in mining revenue in Q226, which will likely worsen in Q326 given BTC's drop from $115 to now $66. However, coupling that massive drop with the massive ramp-up in AI revenue, AI revenue might overtake mining as early as Q326 (should BTC not recover).
- Landing Microsoft seems like only the start. They have 3 additional pockets of Blackwell capacity (Horizon 5-10, Sweetwater 1-2, and retrofits across two BC sites) opening up capacity over the next few years – which seem likely to hook another whale customer or three.
- What matters most now is executing for Microsoft (Horizon 1-4) and proving the operational excellence of their buildout expertise from the start. Simultaneously, they have a multi-year runway from here with their 3 active buildouts/retrofits.