Nvidia halts RTX 5090 production as GB202 chips shift to $15,000 AI data center GPUs

Nvidia Redirects GB202 Chips From Gaming to Professional Markets

Nvidia has reportedly stopped allocating its flagship GB202 graphics processor to consumer GeForce RTX 5090 cards, according to leaked factory notices cited by hardware leakers @hongxing2020 and @Zed__Wang (MEGAsizeGPU).1

1

The decision affects both the standard RTX 5090 and the China-specific RTX 5090 D V2 variant, with all future GB202 supplies now reserved exclusively for RTX Pro AI GPUs and professional visualization workloads.2

2

While Nvidia has not officially confirmed the move, the leaked factory notification states that the company will no longer manufacture any RTX 5090 models going forward.

Economic Logic Behind the Strategic Pivot

The reported shift makes substantial financial sense for Nvidia. The same GB202 silicon that powers the $1,999 RTX 5090 also drives significantly higher-margin AI products in the professional segment. The RTX Pro 5500 features identical 21,760 CUDA cores as the RTX 5090 but comes equipped with 84GB of GDDR7 memory instead of 32GB, targeting AI workloads and professional visualization applications.1

1

The fully-configured RTX Pro 6000, featuring 24,064 CUDA cores and 96GB of memory, sells for approximately $15,000.3

3

By redirecting GB202 allocation toward these professional GPUs, Nvidia prevents its relatively inexpensive consumer graphics cards from competing with its own AI data center accelerators while maximizing returns on limited semiconductor manufacturing capacity.

Supply Shortage Timeline and Price Implications

The RTX 5090 will not disappear from retail shelves immediately. Nvidia's board partners, distributors, and retailers maintain existing inventories of GB202 chips and finished graphics cards that could sustain availability for some time.2

2

However, once current stock depletes, availability will inevitably decline, driving prices higher in an already constrained market. The RTX 5090 launched at $1,999 in January 2025 but has been spotted selling for as much as $9,600 at some retailers, with AI buyers purchasing bulk inventory before cards reach gaming customers.3

3

Regional availability will vary considerably as the supply shortage progresses through different distribution channels.

RTX 5080 24GB Emerges as New Gaming Flagship

According to the leaked factory notice, the RTX 5080 24GB will become Nvidia's highest-end GeForce graphics card, while the RTX 5080 16GB version faces discontinuation.1

1

This rumored RTX 5080 24GB could represent a revival of previously postponed GeForce RTX 50 Super plans, which were delayed due to memory supply constraints. With GB202 chips no longer allocated to gaming products, the new gaming flagship would benefit from improved memory availability. However, questions remain about whether this card will match RTX 5090 performance, especially if it retains the smaller GB203 chip found in the existing 16GB model rather than the larger Blackwell GPU architecture.

Impact on Enthusiast Gaming Market and Future Outlook

The reported end of life for GB202-based gaming cards creates a significant gap in the enthusiast gaming market. The RTX 5090 currently stands as the only GPU capable of handling DLSS 5 neural rendering at 4K without Multi-Frame Generation, making it essential for cutting-edge gaming experiences.3

3

Nvidia's next-generation Rubin-based RTX 60-series GPUs are not expected until late 2027 or possibly 2028, leaving the enthusiast gaming market without a true flagship for an extended period.1

1

Demand from workstation customers and AI developers for professional accelerators will likely continue driving Nvidia's allocation decisions, as these higher-margin AI products generate substantially more revenue than consumer graphics cards. GB202's unusually large die size makes yield and wafer allocation particularly consequential, favoring professional markets where pricing flexibility accommodates manufacturing complexity.