Iren co-founder and co-CEO Daniel Roberts said global computing capacity may never catch up with the growth in artificial intelligence demand. According to Sina Finance, he said the current AI data center buildout is fundamentally different from earlier investment cycles, because each additional unit of compute supply can generate several times more demand.
Roberts said the physical world itself limits oversupply, adding that builders have already hit social, political, and physical thresholds on power supply. He said communities in many parts of the United States are opposing data center construction over energy and water concerns.
Roberts also said Iren chose to own the full chain from land and buildings to computing equipment and software, unlike competitors that use long-term leases to lock in hyperscale customers. He compared those leases to disguised sales and said they had given up what is now the world’s scarcest asset.
Nvidia has played a central role in Iren’s transformation. In May, Nvidia signed a five-year, $3.4 billion compute leasing contract with Iren and gained the right to buy up to $2.1 billion of stock at $70 per share, with the right vesting in tranches as Iren receives up to 600,000 Nvidia chips.
Goldman Sachs expects U.S. data center capacity to double from 2024 levels by the end of 2027 and to more than triple by 2030, reaching about 125 gigawatts. Iren’s customers include Microsoft, Perplexity, and Nvidia.