Nvidia's Sales Suggest Manufacturing Capacity, Not Demand, is the Limit

Nvidia's manufacturing capacity is likely being limited by demand rather than lack of appetite for its products, according to recent data from Oracle. In its fiscal first quarter of 2027, which ended August 31st, Oracle delivered 850 megawatts of AI capacity containing over 300,000 graphics processing units (GPUs) to customers.

Oracle's GPU fleet ran at 97.9% utilization during this period, meaning nearly every chip was busy serving customers. The majority of these chips are Nvidia's, with Oracle's cluster design for its biggest AI deployments connecting hundreds of thousands of Nvidia GPUs.

The revenue opportunity per gigawatt of capacity is also rising. A gigawatt of Nvidia's older Hopper systems meant about $18 billion in revenue opportunity, while the new Vera Rubin platform is worth around $40 billion. This suggests that customers are willing to pay a premium for Nvidia's products.

Oracle's data shows that even older GPUs, some of which date back to 2019, are still commanding a 20% premium when they come up for renewal or resale. This indicates that demand for these chips is outstripping supply, rather than the other way around.