Gaming Takes a Backseat to AI: NVIDIA May Halt Flagship RTX 5090 Production to Focus on AI GPUs

A serious supply shortage is brewing in the high-end PC hardware market. According to multiple reputable Asian insiders, NVIDIA has decided to halt shipments of GB202 GPUs to board partners manufacturing the gaming GeForce RTX 5090. The company plans to redirect the freed-up silicon to its RTX PRO professional accelerators, which target enterprise AI and heavy workstation workloads.

Insider MEGAsizeGPU was first to report the supply freeze, followed by Chinese leaker hongxing2020, who published excerpts from factory notices regarding the halt of chip quota allocations. The leak was corroborated by Taiwan-based BenchLife’s sources among add-in card (AIC) partners. NVIDIA has not yet commented on the situation, meaning the move technically remains unconfirmed, but the chipmaker’s business logic is clear.

Silicon Economics: Why Gamers Are Losing Out to Data Centers

Based on the Blackwell architecture, the GB202 GPU is the most complex and expensive die in the consumer segment, packing 21,760 CUDA cores and paired with 32GB of high-speed GDDR7 memory. The very same silicon powers enterprise products, where profit margins are on an entirely different level.

For instance, the professional RTX PRO 5500 Blackwell accelerator comes with 84GB of GDDR7 memory and sells to enterprise customers at a massive premium for training neural networks and running complex engineering simulations. Amid a global shortage of cutting-edge GDDR7 memory chips, spending scarce silicon on retail gamers simply makes no financial sense for the company.

The flagship’s shortage has already hit retail: as tech outlet Tom’s Hardware reports, despite a $1,999 MSRP, actual prices at US retailers have surged past $5,000, with some scalper listings reaching an astronomical $7,000–$9,000.

A 24GB RTX 5080 as a Stopgap Replacement

To avoid leaving the retail market empty-handed, NVIDIA is reportedly preparing an interim solution. Insiders say an updated GeForce RTX 5080 packing 24GB of GDDR7 VRAM instead of the standard 16GB will arrive in the first quarter of 2027. It could effectively take over as the flagship of the consumer lineup.

However, the card won’t be a true replacement for the top-tier GPU. The RTX 5080 is built around the smaller GB203 die with fewer compute units. While the extra VRAM will help with running local neural networks and demanding 3D workloads, additional memory cannot make up for the physical deficit of CUDA cores in demanding 4K games.

For now, the RTX 5090 officially remains in NVIDIA’s lineup, and previously manufactured batches will trickle into stores for a while longer. Even so, the move sets a clear precedent: the generative AI boom is beginning to directly crowd out top-tier gaming GPUs from the consumer market.