Nvidia Taps IREN for $3.4B GPU Cloud and a $2.1B Share Option, Anchoring a 5GW AI Buildout

The miner-to-AI pivot just got a major endorsement. IREN, once known primarily for Bitcoin mining, secured a multiyear pact with Nvidia that pairs guaranteed GPU cloud spend with strategic equity optionality—an arrangement designed to scale hard power and compute without overcommitting either party to control.

The structure is the tell. Over five years, IREN will deliver $3.4 billion of managed GPU cloud services for Nvidia’s internal AI and research workloads. Alongside the services agreement, Nvidia obtained a five-year option to buy up to 30 million IREN shares at $70 each—potential investment rights worth $2.1 billion. That call option aligns incentives and gives Nvidia upside exposure to IREN’s execution without forcing near-term consolidation, a pattern we’re seeing as hyperscalers and chip leaders secure capacity while preserving flexibility.

On the ground, the partnership centers on deploying Nvidia’s DSX architecture across IREN’s global data center pipeline, targeting up to 5 gigawatts of AI infrastructure. Work begins at IREN’s 2-gigawatt Sweetwater campus in Texas, a site purpose-built for high-density, low-latency GPU clusters. Nvidia’s leadership framed “AI factories” as emerging economic infrastructure that needs tight integration across compute, networking, software, power, and operations—areas where IREN has scaled expertise. IREN’s co-CEO described the tie-up as marrying Nvidia’s systems leadership with IREN’s strengths in power, land, data centers, GPU deployment, and operations.

Capacity is the other hinge. To support the rollout, IREN agreed to acquire Spain-based data center developer Ingenostrum (Nostrum Group), adding 490 megawatts of grid-connected power in Spain. With that, IREN’s total power portfolio reaches 5 gigawatts—deliberately matching the ambition of the Nvidia deployment and helping backstop obligations to other large tech customers.

Markets reacted quickly, then reassessed. IREN shares spiked above $72 in after-hours trading after closing at $56.85, before enthusiasm cooled when the company posted a Q1 net loss of $247.8 million. Recently, the stock was up about 3% on the day at $58.60. Bernstein set a $100 price target following the news. Nvidia, meanwhile, traded above $215 near record levels, up roughly 83% over the past year.

This isn’t IREN’s first big compute commitment. In November 2025, the company announced a $9.7 billion GPU cloud deal with Microsoft, powered by Nvidia GB300 GPUs at IREN’s Childress, Texas facility, including a separate $5.8 billion purchase agreement with Dell Technologies for hardware. With the new Nvidia arrangement, IREN’s contracted and option-linked commitments now exceed $15 billion, positioning the firm as a critical intermediary between chip manufacturers and enterprises chasing dedicated AI compute.

The competitive backdrop is intensifying. Hut 8 signed a $9.8 billion AI data center lease that reset its stock to record highs, and operators like Core Scientific and Terawulf have inked multi-billion-dollar compute agreements with AI platforms. As demand consolidates around a few GPU supply chains, the winners often blend three ingredients: power procurement at scale, rapid DSX-class deployment capability, and balance-sheet structures that keep optionality intact.

What matters next is execution against three friction points: the pace of DSX rollouts at Sweetwater and Europe, the durability of power sourcing in Spain and Texas as grid dynamics evolve, and margin discipline on managed GPU cloud as commitments stack up. If IREN threads that needle, Nvidia’s option could become not just a hedge, but a gateway to deeper alignment—without either side giving up strategic room to maneuver.