Nvidia RTX 5090 Disappears From Shelves as AI Demand Pushes Prices Beyond $9,500

AI Server Builds Drain Consumer GPU Supply

The Nvidia RTX 5090 has effectively disappeared from first-party inventory at major US online retailers including Newegg, Amazon, and Best Buy in less than two weeks1

1

. What remains are listings from third-party sellers demanding between $6,500 and $9,500 for Nvidia's flagship GPU—a staggering increase from its $1,999 launch price in January 20253

3

. The GPU shortage has intensified dramatically, with median prices jumping from $4,299 in June to over $5,199 by early September1

1

. Photographs circulated by Hong Kong outlet HKEPC show wooden pallets stacked with boxed retail GeForce RTX 5090 cards at AI server assembly sites, revealing how consumer GPUs for AI are being deployed en masse2

2

. These aren't workstation GPUs like the RTX PRO 6000—they're the same MSI, Gigabyte, and Asus cards marketed to gamers4

4

.

Why AI Firms Choose Consumer Cards Over Professional Hardware

The economics driving bulk purchases of RTX 5090 cards reveal a compelling value proposition for AI companies. Nvidia's RTX PRO 6000 workstation GPU costs $16,000 and features 96GB of VRAM, but many AI deployments can function with the RTX 5090's 32GB of GDDR7 memory4

4

. The high-end gaming GPU packs the GB202 GPU with nearly identical compute performance to professional cards, making even a $5,000 RTX 5090 attractive for building AI servers1

1

. Jensen Huang recently disclosed that 100,000 Nvidia GPUs trained OpenAI GPT-6 Astra, with plans to quadruple that hardware allocation3

3

. While those likely weren't RTX 5090 cards specifically, the revelation underscores the voracious appetite for GPU compute in AI development. Nvidia's GeForce driver license has prohibited data center deployment since 2017, yet enforcement appears minimal as AI firms continue acquiring consumer hardware2

2

.

Inflated Prices and Scam Risks Plague Remaining Inventory

The cheapest available RTX 5090 on Newegg now comes from Slava Computers at $6,449, while Amazon lists the Asus TUF Gaming OC for $6,395 from Joes Tech Shop—a seller with recent one-star reviews citing unfulfilled orders1

1

. Some third-party sellers push prices even higher, with the MSI Ventus 3X OC listed at $8,699 on Newegg1

1

. European pricing has cleared €5,200, reflecting global AI demand pressures4

4

. Scam activity has intensified alongside inflated prices. In January, 42 Amazon customers received fanny packs instead of RTX 5090 cards in a $999 scam, while recent incidents involved sellers shipping phantom graphics cards with cores and memory removed1

1

. Micro Center remains the sole exception, selling cards in-person around $4,299, insulated from online buyouts through its physical-only sales policy1

1

.

Nvidia's Shifting Priorities Leave Gamers Behind

Nvidia's homepage metadata now lists AI and high-performance computing (HPC) before gaming, signaling the company's transformed identity4

4

. Unlike 2021 when Nvidia implemented Lite Hash Rate restrictions to combat cryptocurrency mining on RTX 30-series cards, no equivalent AI-limiting measures exist for the RTX 50904

4

. The company created an RTX 5090 D variant for China that reduces AI performance to 2,375 AI TOPS compared to the standard model's 3,352 while maintaining gaming capabilities, proving technological restrictions are feasible2

2

. Recent research shows 12.5 million high-end graphics cards shipped last quarter despite rising prices, with Nvidia maintaining a 90% market share3

3

. Features like DLSS 5 Neural Rendering, while enhancing gaming performance, simultaneously maintain the RTX 5090's attractiveness for AI workloads1

1

. The RTX 5080 faces similar pressure, as AI deployments readily accept 16GB of VRAM, effectively pushing realistic consumer options down to the RTX 5070 Ti and below—still priced over $1,0004

4

. Watch for potential production shifts toward professional products and whether Nvidia implements any measures to preserve consumer GPU availability, though current trends suggest premium gaming hardware may become increasingly inaccessible to individual buyers5

5

.