Nvidia Reportedly Redirects RTX 5090 Silicon

Nvidia appears increasingly willing to sacrifice consumer GPU supply when the same silicon can command workstation and AI-infrastructure margins.

• –The RTX PRO 5500 reportedly uses the same 21,760 CUDA-core configuration as the RTX 5090 but offers 84GB of GDDR7 instead of 32GB.
• –Nvidia’s official RTX PRO 6000 page positions its 96GB of memory for local inference, model development, and agentic AI workloads. [Nvidia](https://www.nvidia.com/en-gb/products/workstations/professional-desktop-gpus/rtx-pro-6000/)
• –Developers lose one of the few relatively accessible single-GPU options for running larger local models without paying workstation pricing.
• –Existing RTX 5090 inventory may continue selling, but reduced allocations would likely mean thinner restocks and higher prices rather than an immediate disappearance.
• –The rumor also highlights how AI demand is reshaping the consumer GPU market: identical silicon is more valuable when sold as professional compute hardware.