IREN Limited (NASDAQ: IREN) has announced a multi-year GPU cloud services contract with Microsoft valued at approximately $9.7 billion, which includes a 20% prepayment. The agreement commits IREN to provide access to NVIDIA GB300 GPUs over five years, with a phased deployment through 2026 at its Childress campus (Texas), a 750 MW facility where the company will build new liquid-cooled data centers capable of supporting a combined 200 MW of critical IT load (Horizon phases 1–4).
In parallel, IREN has finalized an agreement with Dell Technologies to acquire GPUs and auxiliary equipment for around $5.8 billion. This figure, according to the company, includes deployment services, servers, InfiniBand, cabling, software, and licenses.
Daniel Roberts, Co-Founder and Co-CEO of IREN: “We are proud to announce this partnership with Microsoft, highlighting the strength and scalability of our fully integrated cloud AI platform. This agreement not only validates IREN as a key provider but also opens a new customer segment among hyperscalers.”
Jonathan Tinter, Vice President of Business Development and Ventures at Microsoft: “Together with IREN, Microsoft is delivering cutting-edge AI infrastructure to our clients. IREN’s experience in building and operating an integrated AI cloud — from data center to GPU stack — combined with its secured energy capacity, makes them a strategic partner.”
What the Contract Includes: NVIDIA GB300 GPUs, phased deployment, and liquid cooling
The announcement positions IREN as a GPU capacity provider to Microsoft over a five-year span, with delivery phases until 2026. The cornerstone of this deployment is the Childress, Texas campus, a 750 MW site where the company will construct liquid-cooled data centers that will collectively provide 200 MW of critical IT load. This liquid-cooled approach reflects sector trends: large-scale AI workloads — training and inferencing of frontier models — demand thermal densities that are hard to sustain with air cooling.
Though IREN does not specify the exact number of accelerators, the reference to NVIDIA GB300 indicates that the new capacity is aimed at intensive training and inference environments with low-latency networks (InfiniBand) and large-scale power supply. The agreement with Dell covers both hardware and the technical logistics required: racks, cabling, licenses, and commissioning services.
How IREN Will Fund the Megaproject
The company explains that it will finance the capex associated through a combination of:
- Available cash.
- Customer prepayments (including the 20% prepayment announced).
- Operational cash flows.
- Additional financing initiatives (to be specified).
This approach is common in hyperscale AI infrastructure deployments, where capacity providers pass on investments to the revenue calendar (take-or-pay contracts, prepayments, multi-year agreements) to limit balance sheet risk.
Childress Campus (Texas): an Energy Hub for the AI Cloud
The Childress site — 750 MW — is a strategic asset in IREN’s portfolio. For AI workloads, power availability, grid quality, and refrigeration are as critical as the GPUs themselves. The company emphasizes that the new liquid-cooled infrastructure will deliver 200 MW of IT load across phases Horizon 1–4, a significant leap for a single location and a showcase of the scale needed for current AI contracts.
It’s no coincidence that the company emphasizes its vertically integrated platform and its energy portfolio: IREN claims to have 3 GW of secured power capacity across North America (US and Canada), in regions rich in renewables. In a time when electricity availability has become a bottleneck for new data centers, access to connected land and energy reserves positions the provider for rapid growth.
What Microsoft Gains and What IREN Risks
For Microsoft, the contract adds an additional supplier of state-of-the-art GPU capacity (GB300) and provides a deployment roadmap through 2026, essential for training and inference of large-scale AI models. For IREN, it validates its role as a large-scale AI cloud service provider, offers multiyear revenue visibility (approximately $9.7 billion total, with 20% prepayment), and access to new hyperscaler clients.
At the same time, the company faces execution challenges: timely closing of financing, ramp-up of liquid-cooled data centers, supply chain management, and coordinated commissioning aligning with Microsoft and Dell schedules.
Why This Contract Is Key for 2026’s “Compute Economy”
The announcement fits into the global race for computational power applied to AI:
- Unprecedented demand for GPUs and lossless network fabrics, supporting training/inference with expanding context windows.
- Thermal transition towards liquid cooling at rack/row scales to sustain densities and efficiency.
- Multi-year procurement and deployment strategies by hyperscalers (contracts of 5–7 years with prepayment and capacity commitments).
Within this context, a $9.7 billion contract centered on GB300 and involving 200 MW of liquid-cooled IT load at a single campus signals that infrastructure — energy, land, data centers, and networks — is as strategic as the chips themselves.
Statements and Roadmap: Insights from the Key Players
Daniel Roberts (IREN) discussed “3 GW integrated portfolio” and a milestone that strengthens the company as a leading AI cloud provider. Jonathan Tinter (Microsoft) highlighted IREN’s experience “from data center to GPU stack” and its secured energy capacity.
IREN added that capacity will be deployed phased until 2026, and expects to finance the project through cash, prepayments, operations, and additional funding. Its agreement with Dell for $5.8 billion covers GPU acquisition and auxiliary equipment (including InfiniBand and licenses), streamlining the supply chain and reducing integration risks.
Risks and Disclaimers: The Fine Print to Remember
As with any major announcement, IREN includes forward-looking statements and a reminder of risks: the ability to execute the strategy, meet revenue targets, develop sites, design/deploy chip-to-chip cooling, and diversify in HPC/colocation. The company refers to its SEC filings for an inventory of risk factors (notably the August 28, 2025 10-K).
Equally important, while the agreement is multi-year and includes prepayment, operational success depends on delivery schedules (chips, networking, racks), electrical grid availability, permits, and coordinated commissioning. Recent history suggests timing will be as critical as megawatt capacity.
Implications for the AI Cloud Map
- For hyperscalers, it increases reliance on dedicated infrastructure partners (like IREN) to accelerate capacity without solely building in-house.
- For hardware suppliers, the agreement with Dell underscores the role of integrators that bundle GPU + networking + software into turnkey AI clusters.
- For energy markets, it highlights the convergence of data centers and grid planning: the 200 MW liquid-cooled load at Childress signals industry scale and importance.
Upcoming Milestones to Watch
- Phase details (Horizon 1–4): commissioning dates and MW allocated per phase.
- Technical profile: rack densities, WUE, and PUE with liquid cooling; InfiniBand topology.
- Financial execution: capex schedule, additional funding structures, and revenue recognition.
- New contracts: whether IREN leverages this milestone to secure more agreements with other hyperscalers.
FAQs
What does the contract between IREN and Microsoft include, and what is its duration?
It’s a multi-year GPU cloud service agreement valued at approximately $9.7 billion, spanning five years, with a 20% prepayment. IREN will provide access to NVIDIA GB300 GPUs deployed in phases until 2026 at its Childress campus (Texas), supporting 200 MW of critical IT load.
How will IREN finance the GPU and data center deployment?
Using existing cash, prepayments from clients, operating cash flows, and additional financing. Simultaneously, IREN has arranged with Dell Technologies the purchase and integration of GPUs and auxiliary equipment for around $5.8 billion (including servers, InfiniBand, cabling, software, and licenses).
What does “200 MW of critical IT load” at Childress mean?
It indicates the usable electrical power dedicated to computing equipment (GPUs, CPUs, networking, storage) within the new liquid-cooled data centers at the campus. It doesn’t represent total plant capacity (750 MW), but the computing load portion during phases Horizon 1–4.
What does Microsoft gain from this deal?
An additional supplier of latest-generation GPU capacity (GB300) and a deployment roadmap through 2026, essential for training and inferencing large AI models. For IREN, this validates its role as a large-scale AI cloud provider, provides multi-year revenue visibility, and grants access to new hyperscaler clients.