$IREN CEO Dan Roberts full interview this morning on CNBC:
• Capex funding
• The bottleneck is bringing GPUs online
• Political pushback on data centers
• Global AI cloud footprint
A few people have Thursday's numbers tangled, so let me untangle them.
The majority of the $684m loss is the cost of retiring Bitcoin miners as we convert those sites to AI Cloud. Non-cash. The cloud business underneath ran ~87% gross margins (ex D&A).
Every megawatt that comes off mining goes back on at multiples of the revenue. Recent 3-year AI Cloud contracts are at >$20m per MW (IT), more than double late last year.
And the $25-30bn of forecasted FY27 capex isn't an equity number. Customer prepayments can cover about half the GPU capex. Lenders can fund most of the rest. We raised ~$19bn over the last twelve months and only ~$3bn of it was equity. We haven't borrowed a dollar against the data centers yet either.
$4bn of ARR is contracted, with three sites commissioning between now and year end to bring it online. And that's the 2026 story. The real ramp is 2027, when Sweetwater and the next wave of capacity come into play.
It's delivery time.
NVIDIA Exemplar Cloud at the compute layer. NVIDIA-Certified Hypervisor via @MirantisIT at the software layer. IREN is the only cloud provider with both.
Execution at scale. From Childress and Sweetwater in Texas to Mackenzie, Prince George and Canal Flats in British Columbia.
IREN is building across multiple campuses simultaneously, with development advancing at Kiowa (Oklahoma), Bundey (Australia) and Badajoz (Spain).
Over 4,000 personnel. Millions of work-hours completed to date. Building at speed and scale.
That is the advantage of IREN's vertically integrated model.
Good news from $IREN On 23 August, MP David Patton was reported to have publicly supported $IREN’s AI campus in Bundy, South Australia. A local councillor also backed the project ahead of public engagement meetings to be held today.
The campus is planned to have a capacity of 800 megawatts and cover an area of around 1,300 hectares.
Estimated investment of around A$10 million.
Target: Start of electricity connection in 2028.
Approximately 500 construction jobs and around 200 permanent jobs.
The site is adjacent to a central power station and the EnergyConnect project infrastructure.
$IREN is planning a closed water system and power supply agreements with local producers. The political and local approach reduces the risks of opposition and permits for $IREN’s first project in Australia. This raises the possibility that the Bondi permit could change the company’s next development in the Asia-Pacific region. However, this is still public support and not final regulatory approval, a customer contract or a new power agreement.
Sources: Coverage in The Australian from 23 August 2026, $IREN’s official Bondi page. Sure. I would put up a hashtag that connects $IREN + Australia + South Australia + AI + data centres + infrastructure + energy, especially since it’s 800MW at Bondi and the news today is about local/political support for the project.
#IREN #IRENLimited #Australia #SouthAustralia #Bundy #AI #Infrastructure #AICloud #DataCenters #DataCenter #Mainframe #GPU #NVIDIA #Hyperscale #Hyperscaler #DigitalInfrastructure #EnergyInfrastructure #CleanEnergy #RenewableEnergy #PowerInfrastructure #EnergyConnection #AsiaPacific #AsiaPacific #TechInfrastructure #AIEconomy #NASDAQ #GrowthStocks $IREN
Congratulations to @IREN_Ltd on achieving NVIDIA Exemplar Cloud status for its NVIDIA GB300 NVL72 deployment.
This validation reflects close engineering collaboration and gives enterprises confidence that demanding AI workloads can perform reliably at scale on IREN’s AI Cloud.
Learn more ⤵️
$IREN | Reminder for the Rest of 2026
• H2 deployment
• H3 deployment
• H4 deployment
• Prince George deployment
• Mackenzie deployment
• Childress deployment
• Remaining 2026 capacity & GPUs to be contracted
• BTC mining fully decommissioned — and removed from quarterly financial reporting
By year-end, I’m expecting IREN to reach a ~$1.1B quarterly revenue run rate as deployments are commissioned, begin billing and AI Cloud revenue ramps.
But what I’m most looking forward to is getting significantly more detail on the long-term AI strategy:
• AI Cloud Strategy — Mirantis + other software/platform integrations and how IREN plans to build out the software layer above its infrastructure.
• NVIDIA DSX Factories — show us the benefits of building and operating datacenters around the DSX architecture, expected costs and expected Token Per MW squeezed out of them.
• Mirantis Product Showcase — actually show us the product & margins. Demonstrate what customers can deploy, how the platform works, the customer experience, orchestration, Kubernetes, GPU management and what capabilities Mirantis brings to IREN's AI Cloud offering.
• Additional Software Solutions — what other software, networking, storage, orchestration, observability and inference solutions does IREN expect to integrate or acquire? More importantly, give us deployment timelines for when these capabilities become available to customers.
• GPU Strategy — procurement, next-generation GPU roadmap, allocation strategy, refresh cycles and how management thinks about GPU depreciation and residual value as new architectures come to market.
• GPU Economics — how the market is currently valuing/renting different GPU generations, payback periods, depreciation, expected useful lives, utilization and how IREN plans to maximize returns before refresh cycles.
• Financing Plan — how the next phase of GPUs and AI infrastructure will be financed while balancing debt, GPU financing, cash and potential equity.
• Grid Capacity Economics — what kind of $/MW economics IREN believes its energized/grid-connected capacity can command as available power becomes increasingly scarce.
• New Sites — I want to hear more about where IREN is looking for its next campuses, how large they could become, expected timelines and what additional capacity could eventually be added beyond the current roadmap.
• Power & Land Acquisition Costs — what does the market look like today for securing new power and land? How much more expensive and difficult is it becoming to acquire land, secure interconnections and bring hundreds of MW online compared with the existing IREN portfolio?
• Replacement Cost of Capacity — what would it actually cost today, and how long would it take, to replicate IREN's existing portfolio of land and grid-connected power? This could be increasingly important when thinking about the underlying value of the infrastructure.
• ERCOT — any updates around interconnection, grid policy, curtailment, power availability and how the changing Texas market impacts the value of IREN's existing capacity.
• Inference Strategy — whether IREN intends to build a broader inference platform/service layer similar to the direction CoreWeave and Nebius have discussed, rather than remaining primarily an infrastructure/GPU provider.
• Technical Roadmap — I would like to hear much more than just "power + land." Talk about networking, storage, orchestration, Kubernetes, scheduling, inference, software, GPU utilization, cluster design, reliability, cooling and how Mirantis & NVIDIA DSX fits into the overall architecture.
That's the part of the IREN story I'm most interested in heading into 2027.
IREN just delivered its first AI data center to Microsoft. Here's what happened.
IREN Limited handed over Horizon 1, a 50MW facility at its Childress, Texas campus, and Microsoft has accepted it. It's the first of four planned 50MW deployments under a five-year, $9.7 billion cloud services contract between the two companies, announced in November 2025.
IREN also achieved NVIDIA Exemplar Cloud status after NVIDIA tested the site's GB300 NVL72 deployment, a designation meant to show the infrastructure can reliably handle demanding AI workloads at scale.
Key numbers:
- Horizon 1: 50MW IT load, direct-to-chip liquid cooled
- Contract: 5-year, $9.7B with Microsoft
- Site team: over 3,000 people worked on delivery
- IREN targets 480MW gross AI Cloud capacity in 2026
- IREN targets 1.2GW gross AI Cloud capacity in 2027
Horizons 2 through 4 are scheduled for delivery later in 2026. IREN's Co-Founder and Co-CEO Daniel Roberts said the milestone shows the company can "execute complex AI infrastructure projects at speed and scale."
Horizon 1: delivered.
IREN has delivered Horizon 1 to Microsoft and achieved NVIDIA Exemplar Cloud status on NVIDIA GB300 NVL72.
Read more: https://t.co/ItVXSXK1V7
AI builders rely on a cloud that moves at their pace. Today, IREN takes another step forward with the completion of our acquisition of Mirantis.
By bringing together IREN’s owned and operated data centers and compute with Mirantis’ flexible, interoperable software layer, we’re giving customers greater choice and control in how they deploy and scale AI workloads.
Developers and enterprises want flexibility as the AI landscape evolves. We're building an open foundation that grows with you, so you’re ready for what's next.
Read more: https://t.co/zxFLgarIGT
The new deep dive on $IREN is now live!👀
I cover everything that has happened the last 4 months, and why the thesis is stronger then ever!🔥
Let me know what u think, and if you find it helpful, please like and share it ;)
Ideas big enough to change the world. Infrastructure built to make them happen.
IREN AI Cloud — What will you build next?
Learn more: https://t.co/P67u9SGbhU