Things are looking up Down Under as Microsoft has made a big, juicy AI deal.
Microsoft has signed a $9.7 billion deal with Sydney-based data center provider IREN, which will provide it with AI cloud capacity for the next five years.
From this deal, Microsoft will gain access to accelerator systems built on Nvidia’s GB300 architecture, designed for AI workloads. As part of the deal, IREN has said it will purchase $5.8 billion worth of GPUs and related equipment.
The deal makes Microsoft IREN’s largest customer, with $1.94 billion in annualized revenue expected over the next five years. Microsoft has already provided a 20 percent prepayment, according to IREN CEO Daniel Roberts.
Demanding times
It is another clear sign of the huge demand for data infrastructure, with even the largest cloud computing providers seeking out other companies to provide them with compute resources. Microsoft signed a similar deal with Nebius, another data center operator, in September, valued at $19.4 billion. That agreement provides Microsoft with AI cloud infrastructure from a data center in Vineland, New Jersey.
IREN is one of a growing list of companies that started as crypto miners and have transitioned successfully to offering data center resources to AI companies. Due to the similar way both operations are managed, with Nvidia GPUs at the forefront, providing resources to AI companies has not been difficult for these miners. CoreWeave and HIVE, two other miners, have also made the successful transition, inking agreements with hyperscalers with tens of billions of dollars.
While Microsoft and others are also pumping tens of billions into building out their own infrastructure, they have been beset by local problems, from community backlash to a lack of reliable electricity and water supply. This has made operators like IREN, which are already operating, very valuable in the interim while demand far exceeds supply. CoreWeave, the largest of these ex-mining operators, has seen its market cap more than triple since debuting on the NASDAQ in March.
Risky business?
There’s growing concern, reflected in the stock market’s recent volatility, that many of these multibillion-dollar investments may not ultimately pay off, as the revenue potential from the current AI boom could fall well short of the trillions expected to be spent over the next decade. But for the infrastructure operators such as Microsoft, there is clearly a substantial gap between the supply and demand at the moment, given its willingness to spend more than $30 billion on these data center deals.
Microsoft is a key player in this latest AI boom not only through Azure, which is the second-largest cloud provider in the world, but through its 27 percent stake in OpenAI. The operator of ChatGPT finally agreed with its key stakeholders and the attorney general’s office of Delaware and California to become a for-profit company, which has opened the doors for a future IPO. Microsoft will continue to own the IP rights to GPT until 2032, and an independent panel will be selected to review when GPT reaches artificial general intelligence. Until that has been reached, Microsoft will receive 20 percent of OpenAI’s revenue, which is expected to be around $12.7 billion this year.
We’re ending on a downer. Last week, the Azure cloud computing platform took down a long list of services from Xbox Live and Microsoft 365 to critical systems for airlines and banks.
