The GeForce RTX 5090 launched with a $1,999 starting price, but finding one anywhere near that level has become increasingly difficult.
Recent US listings have pushed some of the cheapest available cards toward $5,000, while European prices have also climbed beyond €5,000. New images circulating online offer one possible explanation for part of that pressure: companies assembling AI servers are reportedly buying retail RTX 5090 graphics cards in bulk.
The photographs show pallets filled with boxed GeForce RTX 5090 cards waiting to be unpacked and installed into multi GPU AI systems.
These are consumer GeForce models rather than RTX PRO 6000 workstation cards or dedicated data center accelerators.
| Detail | RTX 5090 situation |
|---|---|
| Official launch price | $1,999 |
| Recent US pricing | Around $5,000 for some of the cheapest cards |
| Recent European pricing | Above €5,000 |
| Reported commercial use | AI servers |
| GPU type | Standard GeForce RTX 5090 |
| Previous similar trend | Crypto mining demand for gaming GPUs |
| AI performance | 3,352 AI TOPS for standard RTX 5090 |
| RTX 5090 D AI performance | 2,375 AI TOPS |
AI companies are reportedly buying gaming GPUs in bulk
The latest images reportedly show commercial AI system builders receiving large shipments of RTX 5090 cards.
Instead of using only NVIDIA's dedicated professional or data center products, some companies appear to be building servers around consumer gaming GPUs because the RTX 5090 still offers significant compute performance.
This is not completely new.
In 2025, examples had already appeared of custom AI servers using seven RTX 5090 cards in a single system.
The newer images suggest the practice may now be taking place on a larger scale, with retail cards arriving by the pallet rather than as individual purchases.
That creates additional competition for the same inventory being sold to PC gamers.
The situation resembles the crypto mining boom
The pattern is similar to what happened during the cryptocurrency mining period.
Mining companies bought large quantities of gaming graphics cards because they offered strong compute performance relative to their cost and availability.
Demand became so strong that ordinary buyers struggled to find many GPUs at normal prices.
NVIDIA eventually responded by introducing Lite Hash Rate versions of several RTX 30 Series cards. Those models deliberately reduced Ethereum mining performance without lowering normal gaming performance.
The restrictions were later bypassed, but they temporarily reduced the appeal of some GeForce cards to large mining operations.
AI workloads have now replaced cryptocurrency as another commercial use case for high end gaming hardware.
Standard RTX 5090 remains much faster for AI than RTX 5090 D
NVIDIA has already shown that it can reduce AI capability while keeping most gaming specifications intact.
The RTX 5090 D, created for markets affected by export restrictions, offers lower AI performance than the standard RTX 5090.
The regular model is rated at around 3,352 AI TOPS, while the original RTX 5090 D is listed at approximately 2,375 AI TOPS.
NVIDIA also placed restrictions on multi GPU operation for the D model.
That difference makes the unrestricted RTX 5090 considerably more attractive for companies building AI systems.
Gaming supply could remain under pressure
Bulk commercial demand is unlikely to be the only reason RTX 5090 prices have risen so sharply.
GPU availability, memory costs, board partner pricing, regional supply and reseller behavior can all affect retail prices.
However, large AI deployments create another source of demand for a product that was originally marketed primarily toward enthusiasts and gamers.
When companies buy dozens or hundreds of cards at once, fewer units remain available through normal retail channels.
NVIDIA has not announced plans to restrict commercial purchases of standard GeForce cards or reduce their AI performance.
As long as the RTX 5090 remains useful for both gaming and AI workloads, buyers from both markets will continue competing for the same hardware.
For PC gamers, that means the flagship GeForce card may remain difficult to find at anything close to its original $1,999 launch price.
Discussion (0)
Be the first to comment.