45k user reviews & 1 million votes, over 29k web hosting brands & 85k plans since 2004!
X
Menu
WHTop → 📋 News → IREN Says 2026 AI Capacity Is Largely Sold Out as Pipeline Tops 5GW
IREN Says 2026 AI Capacity Is Largely Sold Out as Pipeline Tops 5GW
📅 - IREN says its 2026 AI infrastructure capacity is largely sold out as the former Bitcoin miner races to expand a data center pipeline exceeding 5GW. The company is targeting roughly 300MW of cumulative IT capacity this year and 800MW in 2027, turning scarce power and GPU access into a fast-growing but capital-heavy hosting business for AI customers.
For infrastructure buyers, the numbers point to a market where access to energized, GPU-ready capacity is becoming commercially distinct from simply owning land or announcing future campuses. IREN says it already has $4 billion of contracted annualized run-rate revenue tied to 2026 capacity, against roughly $1 billion of operating ARR today. [...][... Check source for end of article ...]
Want to add a website news or press release ? Just do it, it's free! Use add web hosting news!
Related news
📅 - IREN Buys Spain’s Nostrum to Enter Europe’s AI Cloud Market - IREN has closed its acquisition of Spain's Nostrum Group, giving the Sydney-based AI cloud operator its first European foothold and roughly 490MW of secured grid-connected power. The deal, whose financial terms were not disclosed, adds a local data center development team and pipeline aimed at customers chasing scarce AI infrastructure capacity across Europe right now.
For infrastructure buyers, the number that matters is not the purchase price. IREN did not disclose it. The number is 490MW. Secured power has become the hard currency of AI infrastructure, particularly in markets where GPUs can be ordered faster than substations can be approved.
IREN, founded in Australia and [...]
📅 - IREN CEO: AI Power Crunch Threatens Gigawatt Data Center Timelines - NVIDIA may not be the AI infrastructure constraint investors think it is. IREN co-founder and co-CEO Daniel Roberts told Bloomberg Tech recently that a company starting a 1-gigawatt AI factory today might wait until 2030 before compute begins operating, reframing the sector's bottleneck around electricity, grid access, permitting, and construction rather than GPU supply alone.
That is a cold shower for a market still trained to read AI capacity through chip availability. NVIDIA remains central, obviously. No one is running frontier training clusters on vibes. But Roberts' remark points to a slower, more physical constraint: land, transformers, substations, transmission queues, power [...]