5GW of NVIDIA DSX Infrastructure Planned in $3.4B IREN Deal

NVIDIA and IREN Limited announced a strategic partnership to deploy up to 5 gigawatts (GW) of next-generation AI infrastructure across a global pipeline. The agreement establishes a massive physical footprint for hardware rollouts while tying the silicon provider directly to grid-scale power assets.

The collaboration centers on NVIDIA’s DSX architecture, a reference design requiring deep integration of accelerated compute, high-speed networking, power systems, and software. IREN provides the physical infrastructure layer, including land acquisition and data center operations, while NVIDIA supplies the silicon and software stack.

Financial Agreements and R&D Compute

The partnership includes two distinct financial mechanisms that bind the companies. IREN signed a five-year, $3.4 billion contract to provide NVIDIA with managed GPU cloud services. This dedicated compute will support NVIDIA’s internal AI research and development.

Simultaneously, NVIDIA secured a five-year warrant to purchase up to 30 million ordinary shares of IREN at an exercise price of $70 per share. If fully exercised, this represents a potential $2.1 billion investment by the chipmaker. IREN’s stock reached approximately $71 in extended trading immediately following the announcement.

Sweetwater Campus and DSX Integration

The deployment pipeline designates IREN’s 2,000-acre Sweetwater campus in Texas as the flagship site for the DSX architecture. The site spans Nolan and Jones Counties and is divided into two operational phases. The 1.4 GW Sweetwater 1 phase was energized the same week as the announcement, with Sweetwater 2 set to add an additional 600 MW.

Deployment Site	Capacity	Assigned Workload	System Specifications
Sweetwater (Phase 1 & 2)	2.0 GW	Flagship DSX Architecture	NVIDIA DSX-aligned
Childress	60 MW	NVIDIA Internal R&D	Air-cooled Blackwell

The internal R&D capacity reserved for NVIDIA will be housed at IREN’s Childress facility. This 60 MW allocation utilizes air-cooled Blackwell systems. Container orchestration for these workloads will be handled in partnership with Mirantis.

The Infrastructure Bottleneck

The scale of this deployment highlights a structural shift in AI engineering. Power procurement and grid interconnects have replaced hardware manufacturing as the primary constraints for AI inference. IREN previously secured a $9.7 billion agreement with Microsoft in November 2025 for GPU cloud infrastructure utilizing NVIDIA GB300 units, cementing its transition from cryptocurrency mining to a dedicated “neocloud” provider.

For enterprise teams managing production AI inference across GPU clusters, these gigawatt-scale facilities change the operational math. When you deploy massive multi-region workloads, physical proximity to dedicated power grids dictates latency limits and scaling caps. You must now factor underlying data center power contracts into your cluster provisioning strategies.