Nvidia has quietly delisted the GeForce RTX 5080 from its own store, following the earlier removal of the RTX 5090, while street prices for the 5080 have climbed to more than 2.2x its original $999 MSRP, Tom's Hardware reported. For teams running local inference or fine-tuning on workstation-class RTX cards, the two removals land as a single supply signal: the parts that matter for GPU-poor shops are the ones disappearing first.
The mechanics, as Tom's Hardware lays them out, are about what stays listed as much as what vanishes. Other Blackwell SKUs — the GeForce RTX 5070 Ti and GeForce RTX 5070 — remain on Nvidia's marketplace, marked "Out of Stock," which the outlet says is the normal pattern: Nvidia keeps pages up through shortages so buyers can watch for restocks. The RTX 5090 and now the RTX 5080 broke that pattern entirely, with their listings and filters removed rather than marked unavailable. Tom's Hardware notes that absence from the store doesn't necessarily mean discontinuation, but it does raise questions about the cards' status and future availability.
The outlet reports rumors that Nvidia has stopped producing the RTX 5090 in order to redirect the GB202 silicon toward the RTX Pro Blackwell lineup, which carries higher margins — a theory Tom's Hardware says has "a lot of credence" given Nvidia's shift toward AI hardware. Parallel rumors, per the report, point to a GeForce RTX 5080 24GB replacing the current 5080 as the new flagship gaming card, addressing the memory complaints about the existing part. Tom's Hardware is explicit that a 24GB 5080 would not be a drop-in substitute for the 5090: the 5080 has roughly half the CUDA cores of the 5090, a gap the outlet says added memory alone can't close.
The number that matters for procurement is the price itself: the cheapest in-stock RTX 5080 Tom's Hardware could find at the time of reporting, a PNY GeForce RTX 5080 ARGB Epic-X RGB OC, was listed at $2,199.99 — 2.2x the card's $999 launch MSRP. The outlet reports that remaining stock is selling out fast and that prices are likely to keep climbing as availability shrinks, with the 5080 unlikely to still be purchasable by Black Friday on current trend. Nvidia has not confirmed discontinuation of either card; Tom's Hardware says it has contacted the company for comment on the 5080's status and had not received a response at publication.
Nothing in the reporting gives a restock date, a production figure, or an official statement from Nvidia explaining either removal — the piece is built entirely on store-listing behavior, retailer pricing, and unconfirmed rumor about the GB202 reallocation and a prospective 24GB variant. That's a thin evidentiary base for anyone trying to plan a hardware refresh: teams can observe the listings disappearing and the prices doubling, but they cannot currently get Nvidia's account of why, or when normal availability might return.
The caveat that matters in week two of any procurement plan built around these cards is that a 24GB RTX 5080, if it materializes, is not a performance substitute for the 5090 it would nominally sit alongside — Tom's Hardware's point about the CUDA core gap means teams waiting for more VRAM at RTX 5080 price points should not expect RTX 5090-class throughput to come with it. Pair that with retailers burning through the last units of the current 5080 at 2.2x MSRP, and the realistic window for buying one at anything resembling launch pricing is already closed.
The takeaway: if your inference or fine-tuning roadmap depends on acquiring RTX 5080- or 5090-class silicon in the next quarter, price in a cloud-GPU fallback now rather than waiting on a restock Nvidia hasn't confirmed is coming.