IREN Recontracts Older Nvidia H100 GPUs at Twice Original Rates

Specialized cloud and data center operator IREN has signed new agreements for older Nvidia H100 graphics processing units at double their original leasing prices, co-founder and co-CEO Daniel Roberts told investors on the company's fiscal first-quarter earnings call.

The renewal pricing indicates that three-year-old AI hardware continues to hold commercial value well past early industry expectations, challenging prevailing assumptions that rapid depreciation cycles would quickly render early generative AI infrastructure uneconomic. IREN had previously assumed in its financial modeling, including on its long-term contract with Microsoft, that GPUs would carry zero residual value after five years of service.

As reported by Yahoo Finance (https://www.tikr.com/blog/iren-ceo-old-nvidia-h100s-just-recontracted-at-double-the-original-contract-price?ref=yahoofinance&tlnk=yf-261106-j3qqpw48&.tsrc=rss) from IREN's Nov. 5 conference call, Roberts pointed to the recontracted H100 fleet as evidence that high-performance compute supply remains constrained across the enterprise market. The company recently signed a five-year contract generating approximately $22 million in annualized revenue per megawatt, alongside a shorter-term agreement priced above $40 million per megawatt, while an unnamed frontier AI research lab also expanded its capacity.

The pricing strength coincides with intense investor scrutiny over IREN's expanding capital requirements. Quarterly capital expenditures jumped to $4.82 billion in the period ending September 2026, up from $281 million during the corresponding quarter in 2025. Short seller Jim Chanos argued on social media platform X that IREN could require roughly $20 billion in average capital employed to reach its revenue ambitions, asserting that a five-to-one ratio of capital to annual revenue was unsustainable over the long term.

Addressing those financing pressures, Roberts noted that the customer on IREN's latest five-year agreement prepaid roughly 90% of the associated GPU capital expenditures, shifting the hardware funding burden away from the operator's balance sheet. Roberts also distinguished short-lived processing chips from IREN's physical data centers, which he characterized as debt-free, 20-year assets currently valued at $6 billion, with a target of $10 billion by June 30.

For the fiscal first quarter, IREN posted revenue of $228 million, up 66% sequentially, driven by AI cloud revenue that more than doubled to $200 million. The company's net loss narrowed to $263 million from $684 million in the prior quarter, while total contracted revenue backlog grew 22% sequentially to $20.3 billion.

Despite the demand, IREN adjusted its operational delivery schedule. Management pushed its target of crossing $4 billion in operational annualized recurring revenue from late 2026 into January or February. Roberts attributed the schedule shift to a customer-directed reconfiguration request from Microsoft at its Childress, Texas, campus to accommodate network infrastructure upgrades for next-generation frontier AI models.