Nvidia has signed a strategic partnership with AI cloud and data center operator IREN to deploy up to 5 GW of DSX-aligned AI infrastructure across the company’s global pipeline. The deal underscores how the AI boom is reshaping the data center market around massive, vertically integrated infrastructure platforms.
The deal positions IREN’s 2 GW Sweetwater campus in Texas as a flagship deployment for Nvidia’s DSX AI factory architecture, a reference design that combines accelerated compute, networking, software, power, and operations for large-scale AI infrastructure.
Nvidia secured a five-year right to purchase up to 30 million shares of IREN stock at $70 per share, creating a potential $2.1 billion investment pathway subject to future conditions and regulatory approvals.
“AI factories are becoming foundational infrastructure for the global economy,” Nvidia founder and CEO Jensen Huang said in a statement. “Deploying these systems at scale requires deep integration across the full stack, compute, networking, software, power and operations.”
The announcement caps a pivotal week for IREN, which has accelerated its transition from bitcoin mining into AI infrastructure and cloud services.
Earlier this week, the company energized the first 1.4 GW phase of its Sweetwater campus and disclosed acquisitions of cloud software company Mirantis and Spain-based infrastructure provider Nostrum Group, adding cloud operations capabilities and nearly 500 MW of European power capacity.
The Nvidia partnership follows IREN’s $625 million acquisition of Mirantis, previously reported by Data Center Knowledge, adding Kubernetes and cloud operations capabilities to help convert deployed GPUs into enterprise-ready AI services.
Together, the moves show IREN assembling nearly every layer needed for the deployment of a hyperscale AI factory, including power, land, data center development, GPU infrastructure, cloud operations, and enterprise AI services.
Texas Emerges as AI Infrastructure Anchor
Nvidia and IREN said future deployments are expected to focus on the Sweetwater campus, which the companies described as a flagship implementation of Nvidia’s DSX architecture.
The partnership also highlights Nvidia’s growing influence over the broader AI infrastructure stack.
“Nvidia’s role is expanding beyond GPUs into full AI infrastructure architecture and orchestration,” said Sameh Boujelbene, vice president at Dell’Oro Group. “The IREN partnership is less about supplying chips and more about deploying up to 5 GW of Nvidia DSX aligned AI infrastructure.”
Rather than simply supplying GPUs to cloud providers, Nvidia increasingly appears to be aligning with operators capable of delivering complete AI factory environments at utility scale.
“The hyperscalers are leaning into their own silicon,” Steven Dickens, CEO and principal analyst at HyperFrame Research, told Data Center Knowledge. “Nvidia has to capture markets outside the hyperscalers, and these neo clouds are racing to raise capital and build infrastructure as quickly as possible.”
Boujelbene said the scale of the announcement reflects a broader shift in how AI infrastructure is being planned and financed.
“A 5 GW AI infrastructure pipeline signals that hyperscale AI demand is moving into a new phase, from opportunistic GPU deployments to multi-year, gigawatt-scale capacity planning,” she said. “It suggests customers expect sustained demand for both training and inference.”
Dickens said former bitcoin mining operators such as IREN hold a major advantage because they already control large amounts of power capacity and data center infrastructure.
“They’ve got pre-built capacity, pre-built data centers, and power,” Dickens said. “That gives them a head start over anybody trying to build this from scratch.”
Many hyperscale cloud campuses historically ranged from a few hundred megawatts to roughly 1 GW of planned capacity. A 5 GW pipeline pushes into territory more commonly associated with regional power infrastructure.
Still, the figure likely reflects long-term deployment potential across IREN’s global portfolio rather than fully committed near-term construction.
Boujelbene said the partnership also reflects a broader move toward vertically integrated AI factory deployments.
“Compute, power, land, cooling, networking, software, and operations are being planned as one system rather than procured separately,” she said. “The next phase of AI infrastructure will be less about buying GPUs alone and more about building repeatable AI factories at gigawatt scale.”
AI Demand Overshadows Weak Quarterly Results
The Nvidia partnership overshadowed weaker quarterly results released alongside the announcement.
IREN reported widening losses and revenue below analyst expectations as the company continues shifting capital and operations away from bitcoin mining toward AI cloud infrastructure and high-density compute deployments.
Investors, however, appeared far more focused on Nvidia’s endorsement and the partnership’s long-term implications.
The deal adds to a broader trend across the AI infrastructure market, where former crypto mining operators are repositioning themselves as AI infrastructure developers by repurposing power-rich sites originally built for energy-intensive digital asset workloads.
“We’re seeing a bunch of firms repurpose their bitcoin mining infrastructure,” Dickens said. “Whoever can put all these pieces together, the power, the infrastructure, the cloud stack, they’ve got a head start.”
That transition is accelerating as hyperscalers and AI model developers search globally for available power capacity, large land parcels, and operators capable of rapidly deploying GPU infrastructure.