IREN inks $2.8 billion in new AI cloud capacity deals and raises annual revenue forecast beyond $4 billion milestone
IREN announced $2.8 billion in new multi-year AI cloud contracts on July 20, with eight customers including Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI, Hume AI, and one unnamed buyer. The deals lift IREN's year-end 2026 revenue target from $3.7 billion to over $4 billion in annualized run-rate revenue, with 85% already under contract.
The substance runs deeper than the headline. Customers have prepaid roughly 45% of GPU capital expenditure tied to their own capacity, with contracts weighted to an average four-year term. That level of prepayment means IREN's customers are bankrolling a significant portion of buildout before hardware deployment—precisely the kind of capital structure AI startups need to avoid drowning in capex.
Five years ago, this trajectory was unimaginable. IREN began as Iris Energy, a renewable-powered Bitcoin miner. After the April 2024 halving, mining economics collapsed, and the stock fell to $5.13 a share. Rather than fail, the company redirected its data centers and power contracts toward GPU capacity instead of mining rigs.
This week's contracts build on a larger pattern. In November 2025, Microsoft signed a $9.7 billion, five-year deal for GPU cloud capacity centered on Nvidia GB300 chips at IREN's Childress, Texas data center. Nvidia followed in May 2026 with a $3.4 billion cloud services agreement for 60 megawatts of capacity at the same location, alongside a strategic partnership covering up to 5 gigawatts of Nvidia DSX-aligned infrastructure. Nvidia also secured a five-year option to purchase up to 30 million IREN shares at $70 each—worth roughly $2.1 billion if exercised. Dell separately agreed to supply approximately $1.6 billion in Blackwell systems for the Nvidia contract. Across Microsoft, Nvidia, and this week's AI developer deals, IREN has announced nearly $16 billion in customer contract value.
Execution risk remains. IREN's targets depend on 480 megawatts of AI Cloud capacity going live by year-end 2026, a substantial undertaking. Revenue only begins ramping once each data center is built, GPUs are commissioned, and customers formally accept them. Timing is critical. IREN has acknowledged that its projections rest on assumptions about GPU models, utilization rates, and pricing.
The customer roster tells the real story. Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI, and Hume AI are not hyperscalers with vast balance sheets—they're AI labs and inference platforms that need immediate GPU capacity without building their own infrastructure. IREN's advantage is clear: hyperscalers have dominated Nvidia supply for years, leaving smaller AI companies scrambling for reserved, affordable clusters.
The broader significance lies in IREN's trajectory. A company that spent years leasing Bitcoin-mining capacity off cheap renewable power has transformed that same infrastructure into a GPU leasing business—same facilities, same cost structure, entirely different market. It now has named, paying customers and a revenue target exceeding $4 billion. Whether the rest of the sector adopts this as a model or treats it as anomalous will reveal how much runway remains in the AI infrastructure buildout.