With the AI wave exploding, almost all computer hardware is now skyrocketing in price, and the RTX 5090 is especially wild. But there’s worse news: NVIDIA may be planning to discontinue this highest-end consumer gaming graphics card and switch to making more profitable workstation products. Recently on X, leaker hongxing2020 posted a screenshot of a Chinese factory notice that read, “5090 EOL, 5080 24GB ready.” (5090 discontinued, 5080 24 GB ready to launch), which set the gaming GPU community ablaze. On the same day, another NVIDIA hardware leaker with a long track record of accuracy, MEGAsizeGPU, followed up and confirmed: “From now on, all GB202 supply will go only to RTX Pro.” The two leaks point to the same thing: NVIDIA may no longer produce the GeForce RTX 5090 and instead give the largest Blackwell desktop chip entirely to its more profitable workstation product line.
5090 EOL🙀，5080 24GB ready 🚀 https://t.co/FUc1igiwlD
— hongxing2020 (@hongxing2020) October 10, 2026
This is not an unfounded GeForce nightmare. The same news was followed up the same day by foreign outlets such as Tom’s Hardware, TechPowerUp, VideoCardz, and HotHardware, while cnBeta and Expreview also carried Chinese-language reports. However, as of now NVIDIA has not officially confirmed the three things—discontinued production, reallocated supply, and new models—nor has it announced any discontinuation date.
The same chip, but the workstation version makes far more money.
The GB202 in the RTX 5090 is NVIDIA’s largest Blackwell desktop chip, with 21,760 CUDA cores and 32GB of GDDR7. In the workstation lineup, the same chip powers the RTX PRO 6000 Blackwell with up to 24,064 CUDA cores and 96GB of GDDR7, and the RTX PRO 5000 series also uses it. The recently unveiled RTX PRO 5500 is even more nuanced: its CUDA core count is exactly the same as the 5090’s at 21,760, while memory goes straight to 84GB, targeting professional visualization and AI applications.
The price gap is the real driving force. The 5090 launched at $1,999 (about NT$65,000, and now sells for over NT$170,000), while a fully enabled RTX PRO 6000 easily sells for $15,000 (about NT$487,500). The industry generally believes NVIDIA probably won’t formally announce EOL, because its approach is simply to steer chips toward more profitable product lines. Discontinuing relatively cheap consumer cards also prevents its own products from cannibalizing its own workstation accelerator cards in the AI market.
The 5080 24GB is taking over, and its specs had already leaked.
The 5080 24GB is not a new name that appeared out of nowhere. As early as September 2024, before Blackwell was announced, VideoCardz reported this configuration, and afterward it was consistently rumored alongside the RTX 5080 SUPER. In July of this year, there were even reports that the entire RTX 50 SUPER series had been sent to board partners, but was held up by GDDR7 pricing.
The current possible specs for a 5080 24GB: switch to 3GB GDDR7 modules, use a GB203-450 chip, higher core clocks, memory speed increased from 30Gbps to 32Gbps, bandwidth from 960GB/s to 1024GB/s, and power consumption from 360W to 415W. Performance would be stronger than the current 16GB version, further widening the gap with the 5070 Ti. Considering the SUPER series is struggling to launch, DRAM makers prefer to produce 3GB modules, and games are demanding more and more memory, this choice is a reasonable solution.
The AI boom is eating up high-end desktop hardware.
The reason NVIDIA discontinued the RTX 5090 is also simple: the AI boom is devouring the supply of high-end desktop hardware. The precedent is Micron, which exited the consumer brand Crucial to focus on data centers, and its stock price surged in response. It’s the same playbook: NVIDIA just swapped memory for GPU chips and Crucial for GeForce.
The reaction to the rumor in the gaming community was blunt. One user compiled the two leaks into a quick summary and ended with a sigh: “Who would’ve thought that being able to buy a 5090 at the original price would actually be a bargain?” Another user, looking at their own 5090, said all they could do now was pray it doesn’t break. Someone else calculated that the entire RTX 50 series rose 66% in a year, and with gamers already caught in a squeeze, the discontinuation rumor is just another blow.
The crossfire from two leakers plus multiple foreign media outlets has pushed the credibility of the 5090 discontinuation rumor beyond ordinary rumor-mongering, but until NVIDIA gives the nod, it remains a rumor (a quiet discontinuation and sales halt is more likely). For gamers, reality is more direct than rumors: the 5090 has long been so expensive it no longer feels like a gaming card, and whether the 5080 24GB can fill the flagship gap—and at what price—are the numbers to watch next. If it is already a foregone conclusion that the workstation line will take GB202, then gaming flagships making way for AI and professional applications will be the norm for the next two years.
Source: KOCPC Chinese