The RTX 50 SUPER Is Not Coming: What to Buy Instead (August 2026)

Bottom Line (August 2026)

• The 24GB RTX 5080 SUPER is not shipping in 2026. Reports have moved from “delayed indefinitely” to rumors of CES 2027. Some coverage suggests an RTX 5080 Ti with 24GB appears instead.
• The cause is memory, not silicon. The refresh needed 3GB GDDR7 modules, which reportedly cost around 3x the 2GB parts. Cards had already reached board partners.
• The RTX 60 series is pushed back too — some reports point to 2028. There is no near-term generation to wait for.
• Waiting costs money in this market. GPU prices rose through 2026. “Wait six months and it gets cheaper” is not how 2026 has worked.
• Buy instead: used RTX 3090 ($1,000-1,300) for 24GB value, RX 7900 XTX ($900-950) for the best new price per gigabyte, Intel Arc B580 ($300-310) under $500, and for 32GB the AMD R9700 ($1,299 MSRP) rather than the Arc Pro B70, which rose to $1,299-1,779 in August and lost its price advantage.

Prices are US street, as of August 2026.

What Actually Happened

The RTX 50 SUPER refresh was a memory-capacity update. The rumored line-up raised VRAM by roughly 50% across the stack: an RTX 5070 SUPER at 18GB, and RTX 5070 Ti SUPER and RTX 5080 SUPER at 24GB. On a fixed 256-bit bus, the only way to get there is denser memory — 3GB GDDR7 modules instead of 2GB.

That is exactly what made it fragile. Reporting through 2026 indicates the cards reached board partners, then the launch went on hold over 3GB GDDR7 pricing, with the denser modules costing roughly three times the 2GB parts. In a year where NVIDIA can sell every GDDR7 die it can get into AI accelerators at far better margins, a gaming refresh whose entire selling point is using more memory per card is the first thing to be cut.

Coverage since has been noisy in both directions — “delayed indefinitely,” then “postponed,” then “back on track” with an RTX 5060 12GB reportedly added, then rumors of CES 2027. Do not plan a purchase around any of those dates. The one thing every version agrees on is that there is no 24GB SUPER to buy in 2026.

Why This Matters More for Local AI Than for Gaming

For gamers the SUPER refresh was a nice-to-have. For local inference it was the most interesting card of the year, because VRAM capacity — not shader performance — decides what models you can run.

A 24GB RTX 5080 SUPER would have been the first new, warrantied, widely-available 24GB NVIDIA card at a non-workstation price in years. It would have made the used RTX 3090 recommendation obsolete overnight. Instead, the 24GB tier still runs on five-year-old used cards, and the cheapest new NVIDIA route to 24GB+ remains the $4,300-5,000 RTX 5090.

The Waiting Trap

Deferring hardware purchases is normally sensible. In 2026 it has been expensive.

NVIDIA has run multiple GeForce price increases this year, street prices sit well above MSRP across the stack, and cards that should be cheap with age are not: the RTX 4070 Ti Super and the RTX 3060 both trade above their own launch MSRPs. The usual assumption — that last-generation cards get cheaper — is inverted right now.

So “wait for the SUPER” has meant paying more for the card you eventually buy, while not having a machine in the meantime. If you genuinely do not need a GPU until 2027, waiting is fine. If you have been deferring a purchase you need, the rumor you are waiting on has already slipped twice.

What to Buy Instead, by Tier

24GBEVGA RTX 3090 24GB ↗

24GBXFX Radeon RX 7900 XTX 24GB ↗

32GBASRock Intel Arc Pro B70 Creator 32GB ↗

The Uncomfortable Conclusion

The SUPER delay is a symptom, not an event. Memory is the scarce resource in computing right now, VRAM capacity is what local AI buyers need, and NVIDIA’s incentive is to route every dense memory module to data-center products instead.

The practical consequence: do not expect the consumer GPU market to solve your VRAM problem in the next year. If you need more than 24GB, the routes that exist today — two used 3090s, a used workstation card, or 128GB of unified memory at low bandwidth — are the routes that will still exist next year. Plan around them rather than around a launch date.

One honest counterpoint: if an RTX 5080 Ti with 24GB does appear at Gamescom or CES, it lands into a market where a used 3090 costs $1,000-1,300, and it would reset the 24GB tier immediately. That is a real possibility, just not one to leave a machine unbuilt over.

See Also

• Best Local LLM for the RTX 5090 — the only fast card above 24GB today
• RTX 5090 vs 4090 vs Used 3090 — the three-way comparison at current prices
• Best GPU Under $500 for Local AI — why the sub-$500 bracket is a 12GB bracket now
• The Cheapest 32GB VRAM GPU — four routes to 32GB, including the Arc Pro B70
• Best 48GB VRAM Setup — where to go when 24GB is not enough
• Should You Buy RAM Now, or Wait Out the Shortage? — the same waiting question for system memory
• Should I Wait for the RTX 60 Series? — how long the next real refresh actually is