NVIDIA May Scrap RTX 5090 Production To Feed Surging AI Demand
Gamers may soon lose access to NVIDIA's flagship graphics card for consumers. According to a pair of prominent leakers on social media platform X, Team Green is preparing to wind down allocations for the GeForce RTX 5090 to redirect GB202 GPU dies into its far more lucrative AI and workstation-grade enterprise lineups. If true, the pivot underscores just how heavily the ongoing AI boom is cannibalizing high-end desktop hardware supplies.
To be fair, that wasn't really in question, regardless of whether the latest unofficial chatter is true. For one, we've already seen companies pivot in the same manner—most notably, Micron wound down its consumer-facing Crucial brand to focus on the data center, and its share price has skyrocketed since then.
"The AI-driven growth in the data center has led to a surge in demand for memory and storage. Micron has made the difficult decision to exit the Crucial consumer business in order to improve supply and support for our larger, strategic customers in faster-growing segments," Micron's Sumit Sadana candidly stated at the time.
As for NVIDIA specifically, its GeForce RTX 5090 has already been an extremely hot commodity. A quick peek at Best Buy's inventory, for example, reveals that nearly every partner model is sold out, as well as NVIDIA's own Founders Edition model. Of the four options that are in stock, prices start at $8,044.99 and scale to over $9,000, all far removed from NVIDIA's $1,999 launch MSRP.
All future GB202 supplies are RTX Pro exclusive
— MEGAsizeGPU (@Zed__Wang) October 10, 2026
So, what about the rumor of NVIDIA effectively discontinuing the GeForce RTX 5090? No such announcement has been made, but according to MEGAsizeGPU on X, NVIDIA is allocating all future GB202 chip supplies to its professional RTX graphic cards.
The GB202 is a big chip based on NVIDIA's Blackwell GPU architecture. It powers the GeForce RTX 5090 with 21,760 CUDA cores, along with multiple RTX Pro models, including the RTX Pro 6000 Blackwell with 24,064 CUDA cores and a whopping 96GB of GDDR6 memory.
Supplying those dies to consumer gaming cards at consumer price points yields thinner margins than packaging them into high-margin enterprise solutions, at least on paper. In practice, it's hard to tell with price hikes affecting the landscape, though it's a safe bet that the data center is still way more profitable. By shifting GB202 production exclusively over to its professional-tier Blackwell cards, NVIDIA can squeeze more revenue out of every wafer coming off TSMC's production lines.
Another leaker, hongxing2020, echoed the sentiment in a separate post saying, "5090 EOL, 5080 24GB ready." A screenshot posted alongside claims NVIDIA has already put the word out to at least one factory that it is no longer selling GeForce RTX 5090 models, including D variants, as spotted and translated by Videocardz.
The idea of a GeForce RTX 5080 with 24GB ties in with past rumors of a Super refresh, though whether it still materializes, we'll have to wait and see.
