Daniel Roberts, co-founder and co-CEO of AI cloud provider Iren, has stated that global compute capacity may never catch up with the surging demand for artificial intelligence. He emphasized that the current wave of AI data center construction is "fundamentally different" from previous investment cycles, as each additional unit of compute supply generates several times more incremental demand.

Roberts founded Iren with his brother in Sydney in 2018, initially as a bitcoin mining operation before pivoting to become an AI infrastructure provider. The company's market value has climbed from a low of roughly $60 million following the crypto market collapse in late 2022 to approximately $17 billion today. Over the past 12 months, Iren has raised about $19 billion through convertible bonds, chip-backed debt, customer prepayments, and equity, with most proceeds directed toward purchasing Nvidia processors and building data centers.

The company plans to invest up to $30 billion in AI infrastructure during the fiscal year ending June 2027. Roberts argued that the physical world itself acts as a natural check on oversupply, noting that builders are already hitting "social, political, and physical thresholds" in power availability. Communities across the United States are increasingly opposing data center projects over concerns about energy and water consumption.

Unlike competitors that lock in hyperscale clients through long-term leases, Iren has chosen to own the full stack—from land and buildings to computing equipment and software layers—to capture the upside from rising compute prices. Roberts likened rivals' long-term lease agreements to a disguised form of selling, suggesting they have "given up the most scarce asset in the world today."

Nvidia has played a central role in Iren's transformation. In May, Nvidia signed a five-year compute rental contract with Iren valued at $3.4 billion, along with an option to purchase up to $2.1 billion in Iren shares at $70 each, vesting in stages as Iren receives up to 600,000 Nvidia chips. Roberts noted that the two companies communicate nearly every day on various matters, including financial issues, with Nvidia continuously supporting all aspects of the ecosystem.

Goldman Sachs projects that U.S. data center capacity will double by the end of 2027 compared to 2024 levels, and more than triple by 2030 to reach roughly 125 gigawatts. Nasdaq-listed Iren counts Microsoft, Perplexity, and Nvidia among its clients. As of the end of June, the company held $7.6 billion in cash along with an equivalent amount of debt, and still has $3.5 billion in share issuance capacity available. Roberts recently remarked on social media that the real acceleration will arrive next year, adding, "Now is the time for delivery."