The RTX 5090 Now Costs $9,400 in the US — 370% Over MSRP, and November Hikes Are Already Leaking

中文版 Chinese version

Prices below are a snapshot from TechSpot’s November 2, 2026 survey and November 3–4 reporting on further hikes, not NVIDIA list prices. Retail prices move fast in this market and are subject to change.

The number that stopped the thread

TechSpot’s Q4 2026 GPU pricing update, published November 2, tracks the cheapest in-stock price for each current graphics card across 10 countries, then compares it with launch MSRP. The headline numbers are brutal:

• RTX 5090 in the US: $9,400, the cheapest in-stock unit — 370% above its $1,999 launch MSRP, up from $6,890 in September, a 36% jump in a single month. Across all 10 countries, the normalized average is about $7,240, or 262% above MSRP.
• GPU prices overall: up 11% in October alone, and about 35% over the past three months. Cards now average 44% above MSRP across the 10 countries, and 58% above in the US and Canada.
• The pain is not confined to the flagship. In the US, the RTX 5080 sits about 120% above MSRP and the RTX 5060 Ti 8GB about 40% above it.

For scale: the cheapest US RTX 5090 in February 2026 was already $3,500. In eight months it has nearly tripled again. A single card now costs roughly what 4.7 cards cost at launch.

Then November’s hikes leaked on top

As if the survey were not enough, November 3–4 reporting pointed to a further round of RTX 50 price increases, attributed to supply-chain and board-partner sources and flagged here as a report/leak, not a confirmed NVIDIA price list:

• RTX 5090D v2: +$330–$500
• RTX 5080: +$380–$450
• RTX 5070 Ti: +$150
• RTX 5060: +$30

Retail channels typically take a week or two to pass wholesale adjustments through to storefront listings, so these numbers, if accurate, are a preview of December shelf prices rather than today’s tags. But the direction is the same as the survey’s, and that is the point: there is no plateau in this data, only acceleration.

Is the 5090 even a gaming card anymore?

Here the story splits into a confirmed fact and a disputed claim, and keeping them apart matters. NVIDIA has denied discontinuing the RTX 5090. What TechSpot found at retail tells a different story in practice: the card is nearly impossible to find through normal retailers or even NVIDIA’s own marketplace, which is why TechSpot concluded it has effectively disappeared for gamers. A leaked factory notice dated October 10 reportedly claimed production had stopped and the GB202 chip was being reserved for professional and data-center products — that notice is unconfirmed, and NVIDIA’s denial stands against it.

The honest summary is: officially still a product, practically a ghost. Gamers are not arguing about a spec sheet; they are arguing about a card they cannot buy at anything near the price NVIDIA announced.

Memory explains part of this — and only part

The standard explanation is real as far as it goes. AI data centers are absorbing DRAM and GDDR supply, memory makers are shifting capacity toward server DRAM and high-bandwidth memory, and TrendForce reported in late September that GDDR6 supply was constrained as manufacturers cut production, with GDDR7 constrained too. That raises the bill of materials for every card.

But it does not explain the whole curve. Among three GeForce cards that all carry 16GB of GDDR7, the cheapest US price exceeds MSRP by anywhere from $361 to $1,201. Same memory capacity, wildly different premiums. That gap is not a memory bill — it is scarcity pricing, allocation, and whatever margin each link in a starved chain can extract. Memory is the trigger; the market structure is the amplifier.

The judgment

The GPU market has quietly changed categories. A $9,400 “gaming” card is not a premium product — it is a professional/data-center asset that happens to have a GeForce badge, priced for buyers who do not flinch at five figures. Everyone else is being repriced out of the high end entirely, and the November leak suggests the mid-range is next in line for the same treatment, just in smaller increments.

The practical advice writes itself, and it is grim: if you do not need a card right now, do not chase this market; if you do, the only defensible move is the cheapest in-stock card that meets a real workload, not the flagship. And the question worth watching is not when prices “normalize” — no survey in this cycle offers a timetable — but whether NVIDIA ever restores the 5090 as a purchasable gaming product, or quietly lets the GeForce flagship become a data-center part with a gaming name.

Sources

• TechSpot, Q4 2026 GPU pricing update, November 2, 2026, as reported by TweakTown, November 3, 2026: https://www.tweaktown.com/news/114468/nvidia-rtx-5090-now-262-percent-above-msrp-globally-as-gpu-prices-jump-35-percent-in-three-months/index.html
• HotHardware, “GeForce RTX 50 GPUs Tipped For Big Price Hikes This Month,” November 4, 2026 (November hike figures attributed to ChannelGate/supply-chain sources; treated here as a leak, not confirmed pricing): https://hothardware.com/news/geforce-rtx-50-gpus-tipped-for-big-price-hikes-this-month
• ShiftDelete, “RTX 5090 Price Hits $9,400, 370% Above MSRP in the US,” November 2026 (summarizing TechSpot’s survey): https://shiftdelete.net/en/rtx-5090-price