Microsoft Reportedly Shifts Away From Older Nvidia GPUs as AI Infrastructure Demand Evolves

Microsoft ($MSFT) is reportedly winding down older Nvidia ($NVDA) computing capacity rented from Oracle ($ORCL) as it reviews billions of dollars in cloud agreements. However, this appears more like a shift toward newer-generation GPUs than a broader retreat from AI spending.
Microsoft recently asked IREN ($IREN) to upgrade its GB300 network to support frontier AI models, while IREN reportedly re-contracted older H100 GPUs at twice their original price.
These developments suggest AI computing demand is shifting across GPU generations, rather than simply disappearing.