Table of Contents
Oracle delivered 850 megawatts of AI capacity containing more than 300,000 GPUs to customers during its fiscal first quarter
In sum – what to know:
850 MW delivered – Oracle delivered 850 megawatts of AI capacity containing more than 300,000 GPUs to customers in Q1.
Funding model evolves – Oracle closed more than $30 billion of additional AI contracts without requiring additional capital from Oracle.
Power becomes critical – Oracle says infrastructure constraints have shifted from GPUs and fabs to power generation and data-center capacity.
Oracle is rapidly expanding its AI infrastructure capacity while using customer prepayments, supplier financing, and other funding models to reduce the amount of capital it needs to provide directly.
Oracle delivered 850 megawatts of AI capacity containing more than 300,000 GPUs to customers during its fiscal first quarter, co-CEO Clay Magouyrk said during the company’s earnings call. The delivery was almost three times what Oracle delivered in the previous quarter and represented 73% of the total capacity delivered during fiscal 2026.
Oracle also said it closed more than $30 billion of additional AI contracts in Q1 without requiring additional capital from Oracle, Magouyrk said.
The distinction between infrastructure investment and Oracle-funded capital was a recurring theme during the call. CFO Hilary Maxson said the majority of new RPO contracts were structured through prepayments, bring-your-own-hardware arrangements or similar mechanisms.
Magouyrk said Oracle has developed several ways to fund the expansion, including supplier financing arrangements, customer-funded hardware, and customer prepayments.
“I think that we have to separate out in our minds what Oracle spends as capex, uncouple that directly from how we think about how the business can grow,” Magouyrk said.
“Because from our perspective, I think we see ways that clearly capital is still required to do this work. It does not all have to flow from Oracle side. It does not have to be Oracle capex,” the executive added.
Oracle expects total capex of $90 billion to $95 billion for fiscal 2027, with no more than $70 billion in net cash capex, according to Maxson.
At the same time, the company is seeing high utilization of the AI infrastructure it has already deployed. GPU utilization reached 97.9% in Q1, while capacity coming up for renewal was renewed or resold at a 20% premium to prior contracts. Magouyrk said the majority of those GPUs were four years or older.
The infrastructure constraints around AI are also changing, according to Magouyrk. “It used to be that the constraints were GPUs and fabs. Then constraints moved to power generation,” he said.
Oracle is developing multiple large-scale data-center sites as it works to bring additional capacity online. Magouyrk said New Mexico and Wisconsin are important sites, but emphasized that Oracle has a broader set of data-center developments in the U.S. and around the world.
In New Mexico, the data center is on track, and Oracle is going through the process of acquiring an air permit. The company plans to use Bloom fuel cells for onsite power generation.
In Wisconsin, Oracle is working with partners to deliver the site’s energy capability through the grid. Magouyrk said data-center delivery there is also on track.
Oracle’s expansion is also being supported by strong demand. Maxson said Oracle expects around half of its RPO to convert into sales over the next 36 months, while Magouyrk said the company remains confident in its ability to meet current and future RPO growth.
