Nvidia reportedly kills the RTX 5090 to force AI datacenters to buy Pro GPUs
Nvidia is reportedly halting production of its top gaming GPU to stop AI builders from using it as a cheap alternative to enterprise chips.
Business · Source: Video Games Chronicle
What happened
Nvidia has reportedly stopped manufacturing its flagship GeForce RTX 5090 graphics card. According to hardware leakers and an alleged factory notice, production of all RTX 5090 models has ceased entirely. This halt reportedly includes the D variants specifically made for the Chinese market. Nvidia has not officially confirmed the move and has ignored requests for comment from hardware publications.
The RTX 5090 launched in January 2025 with a retail price of $1,999. It currently remains the absolute best graphics card on UserBenchmark in terms of raw performance. However, street prices quickly skyrocketed to over $4,000 or even $5,000. Retailers sell out extremely quickly when new stock arrives because AI companies buy up the inventory. The core issue is the GB202 silicon used inside the RTX 5090. This exact same silicon is also used in Nvidia's RTX Pro GPUs.
Those RTX Pro cards are significantly more expensive and are built specifically for AI data centers and processor-intensive environments. With the RTX 5090 out of the picture, the RTX 5080 24GB will reportedly become Nvidia's highest-end consumer graphics card. The 16GB version of the RTX 5080 is also reportedly being discontinued. Meanwhile, the next-generation Rubin-based RTX 60-series is not expected to arrive until late 2027 or possibly 2028.
Key facts
- $1,999 — Original launch price of the RTX 5090 in January 2025
- $4,000-$5,000 — Current street price range for the RTX 5090
- GB202 — The silicon chip shared by the RTX 5090 and RTX Pro GPUs
- 2027 — Earliest expected arrival for Rubin-based RTX 60-series
Why it matters
Hardware arbitrage is getting shut down completely. Builders and smaller AI data centers have been buying consumer RTX 5090s to train neural networks and run parallel processing tasks on a tight budget. GPUs perform technical calculations faster and with greater energy efficiency than CPUs. Now, those teams will be forced to buy significantly more expensive RTX Pro GPUs to get access to the exact same GB202 silicon. Nvidia is effectively closing a massive loophole that allowed AI startups to avoid paying enterprise hardware premiums.
This signals a permanent shift in how hardware companies view their product lines and consumer bases. Consumer gaming is no longer the priority when massive AI margins are on the line. Because Nvidia faces zero direct competition from AMD at the highest end of the performance market, it can safely sacrifice its best gaming card without losing its dominant position in PC gaming. Expect the remaining RTX 5090 stock to see massive price spikes on the secondary market as desperate buyers fight over the last available units.
For builders
Compute costs rise for AI startups
Smaller AI teams relying on consumer hardware will lose access to the most powerful budget-friendly chip. Startups will pay a massive premium to secure RTX Pro GPUs for their data centers. Nvidia wins the margin game, while bootstrapped founders lose their best compute option.
Secondary market prices will explode
Remaining RTX 5090 inventory is already selling between $4,000 and $5,000. Hardware flippers will win big as supply dries up completely. Developers needing local high-end compute will pay extortionate street prices just to keep building.
Hardware strategy shifts to memory limits
Nvidia is reportedly delaying its RTX 50 Super refresh due to severe memory supply constraints. Builders planning local hardware clusters must account for industry-wide memory shortages. This same shortage is reportedly delaying the PlayStation 6 and impacting Xbox hardware pricing.
My take
The RTX 5090 was not killed by a sudden chip shortage or manufacturing defect. I look at this and see Nvidia protecting its own profit margins because a $1,999 gaming card was cannibalizing its enterprise AI business. When your consumer product becomes a direct threat to your massive data center revenue, you kill the consumer product.
Original reporting: Video Games Chronicle. This is my rewrite and opinion.