Microsoft inks $9.7B AI cloud deal with IREN to lock in Nvidia GB300 capacity

Microsoft has agreed to a five-year, $9.7 billion cloud capacity contract with IREN, securing access to Nvidia’s GB300 GPUs as the company races to meet surging AI demand without building new data centers from scratch. The deal includes phased deployments through 2026 at IREN’s 750 MW Childress, Texas campus.

The Breakthrough

• $9.7B, five-year agreement gives Microsoft dedicated access to Nvidia GB300 systems via IREN.
• IREN will add liquid‑cooled data centers delivering about 200 MW of critical IT capacity at the site.
• Separate $5.8B IREN deal with Dell will supply Nvidia hardware and infrastructure.
• IREN shares jumped more than 20% on the news; Microsoft addresses ongoing compute constraints.

Technical Details

• Hardware: Nvidia GB300 GPUs; equipment sourced via a $5.8B agreement with Dell Technologies.
• Deployment: Phased rollouts through 2026 at the Childress campus, with new liquid‑cooled facilities.
• Power and scale: 200 MW critical IT load at Childress; IREN operates 2,910 MW of data-center capacity across North America.
• Commercial terms: Contract includes a 20% prepayment from Microsoft, aiding IREN’s capex.

Impact and Applications

• For Microsoft Azure, the capacity adds near‑term headroom for AI services like model training and inference without incurring heavy up‑front capex or new power procurements.
• The agreement underscores the rise of “neocloud” providers—specialists scaling Nvidia-based clusters for hyperscalers—following Microsoft’s separate multiyear Nebius deal worth up to $17.4B.
• Market reaction: IREN’s stock spiked 20%+ intraday as analysts flagged the contract as “game‑changing” for the operator’s AI pivot.

Future Outlook

• Timeline: GB300 systems to be delivered and installed in phases through 2026; contract depends on IREN meeting milestones.
• Constraint relief: Microsoft has warned AI infrastructure shortages could stretch into mid‑2026—this deal is positioned to narrow that gap.
• Watch factors: execution risk on liquid‑cooling buildouts, supply timing for GB300s, and financing tied to prepayments and equipment purchases.

In short, locking in GB300 capacity via IREN gives Microsoft a faster route to scalable, liquid‑cooled AI compute, while validating a growing ecosystem of specialized infrastructure partners built around Nvidia’s latest chips.