NVIDIA in 2026: No New Gaming GPU, AI Takes Over

For roughly three decades, Nvidia has launched a new GeForce generation about every 18 to 24 months, usually with a mid-cycle Super or Ti refresh in between. Gamers learned the rhythm: wait for the new architecture, compare the benchmarks, buy the card that fits your budget, and repeat two years later.

In 2026, that rhythm broke. According to multiple industry reports tracking Nvidia's release cadence, Nvidia will not ship a single new gaming GPU in 2026 — not a new architecture, not a refresh, not a budget card. If that holds, it will be the first calendar year without a new GeForce launch since the original GeForce 256 shipped in October 1999.

That's a striking claim, so it's worth being precise about what is confirmed, what is reported, and what is still just a rumour. Because the difference matters if you're deciding whether to buy now or wait.

What is actually confirmed — and what isn't

Here's the part Nvidia has not disputed in a way that changes the story: there is no announced RTX 60 series, no announced RTX 50 Super refresh on shelves, and no new gaming architecture launched in 2026. The RTX 50 series, which launched across 2025, remains the current GeForce lineup.

What Nvidia has said on the record, in a statement to Tom's Hardware reported in February 2026 coverage, is that demand for GeForce RTX GPUs is strong, and memory supply is constrained. It did not directly address the delay claims, and it did not announce a new gaming GPU to contradict them. That silence, combined with the lack of any launch, is why so many outlets have converged on the same conclusion: no new gaming GPU in 2026.

What is not confirmed by Nvidia itself is the exact RTX 60 date. The timeline has slipped repeatedly in leak-based reporting, and leaks slip for a reason — they're not announcements. Here's how that timeline has moved, according to reporting that tracks leakers such as kopite7kimi and Kepler_L2:

• January 2026: RTX 60 expected in the second half of 2027.
• February 2026: The Information, via Tom's Hardware, reported no new gaming GPUs in 2026, and RTX 60 mass production pushed past late 2027.
• September 2026: Multiple leakers converged on 2028.
• October 2026: Tom's Hardware reporting put the window at late 2027 at the earliest, or 2028.

Treat 2028 as the current best guess from leaks, not as a date Nvidia has promised. If you're planning a build around an RTX 60 launch, you're planning around a rumour with a history of slipping.

Why would Nvidia skip gaming for a year?

The short answer, from reporting across the industry, is memory — and where the money is.

A severe global memory shortage has forced hard choices about where limited memory goes. AI accelerators, which sell at much higher margins and in much larger volumes to cloud providers, get priority. Gaming GPUs, which need the same memory, get less of it. One third-party analysis, AInvest, described a planned 30–40% cut in GeForce production in early 2026, with some models pulled from the market, and framed it as a supply-driven reallocation, not a demand collapse. Nvidia has not confirmed those specific cut percentages, so treat them as reported analysis, not as Nvidia's own figures.

The broader financial picture, also from third-party reporting, is that gaming is now a much smaller share of Nvidia's business than it was. AInvest's analysis put gaming at about 8% of total revenue in the nine months to October, down from 35% in 2022, with data centre revenue at $51.2 billion in the latest quarter, representing 88% of total sales. Again, those are reported figures to check against Nvidia's own filings if you're making investment decisions — but the direction is clear even without the exact percentages: Nvidia is now, first and foremost, an AI infrastructure company that also makes gaming GPUs, not the other way around.

In 2026, that has meant no new gaming launch at all.

Meanwhile, in AI: Rubin CPX

While gaming waited, Nvidia's data centre roadmap moved. And the most interesting new chip of 2026 for AI isn't a gaming card at all — it's Rubin CPX.

Announced in October 2026, Rubin CPX is a new class of GPU purpose-built for large-scale inference, specifically for massive-context processing. In Nvidia's words, it's the first CUDA GPU purpose-built for massive-context AI, where models reason across millions of tokens of knowledge at once.

The headline specs, from Data Center Dynamics' report of the announcement, are:

• Up to 30 petaflops of AI compute in NVFP4 format.
• 128GB of GDDR7 memory.
• Integrated video decoders and encoders, plus long-context inference processing, in a single chip.
• Availability expected at the end of 2026.

Rubin CPX is designed to sit alongside Vera CPUs and Rubin GPUs inside the Vera Rubin NVL144 CPX rack-scale platform. That rack, a liquid-cooled integrated system, is claimed to offer eight exaflops of AI compute in NVFP4, 7.5 times more AI performance than GB300 NVL72 systems, with 100TB of fast memory and 1.7 petabytes per second of memory bandwidth in a single rack. A dual-rack solution, combining an NVL144 CPX rack with a regular Vera Rubin NVL144 rack, is also planned.

Early adopters named in the announcement coverage include coding tool Cursor, generative AI firm Runway, and AI firm Magic — all companies where long-context video, code, or generative work benefits from a chip built for inference at scale, rather than just for training.

Nvidia founder and CEO Jensen Huang framed CPX as a new category of processor, not just a faster version of an existing one. Whether that framing holds up in real deployments will depend on software, pricing, and how quickly cloud providers actually offer CPX capacity. But the intent is clear: Nvidia is building specialised silicon for the inference era, where the money is increasingly in running models all day, every day, not just in training them once.

The wider Rubin roadmap

Rubin CPX doesn't exist in isolation. It's part of a data centre sequence Nvidia set out at GTC in March 2025, and reaffirmed in January 2026:

• Blackwell Ultra in 2025.
• Vera Rubin in 2026.
• Rubin Ultra in 2027.
• Feynman in 2028.

Those are data centre plans, not GeForce launch dates, and it's important not to confuse the two. A roadmap year is not a delivery guarantee, and an announcement is not the same as production shipment, which is not the same as usable cloud deployment.

On that last point, there is real progress to report. Nvidia CFO Colette Kress said on 26 August 2026 that production shipments of Vera Rubin began earlier that month. CoreWeave announced availability on its cloud on 30 September, which Nvidia described as early access. So Rubin is moving from roadmap to reality, even while gaming GPUs stand still.

For anyone buying AI infrastructure, the practical lesson is to plan contract reviews around that sequence, but to verify availability with your cloud provider before you commit. For gamers, the lesson is simpler and less welcome: the engineering talent, the memory, and the fab capacity are going to AI first, and gaming will get its turn when the supply and roadmap allow.

What about DLSS 5?

If you're a gamer, there's one software bright spot in 2026: DLSS 5.

DLSS 5 launched on 3 September 2026, and for now it runs only on RTX 50 cards, with RTX 40 support planned but undated, according to tracking of the launch. That means owners of current-generation cards get a meaningful upgrade without buying new hardware, while owners of older cards are left waiting for a date Nvidia hasn't given yet.

DLSS matters because it changes what "performance" means. Instead of raw rasterisation power alone, modern gaming performance is increasingly about upscaling, frame generation, and AI-assisted rendering. A software upgrade that improves frame rates and image quality on existing hardware is genuinely valuable — but it's not a substitute for a new GPU generation, and it doesn't help you if you're still on a card that DLSS 5 doesn't support yet.

If you already own an RTX 50 card, DLSS 5 is a reason to feel better about your purchase in a year with no new launch. If you don't, it's a reason to factor DLSS support into your buying decision, not to wait indefinitely for a card that may not arrive until 2028.

So, what should you buy now?

This is the question most readers actually care about, so let's be direct.

If you need a graphics card now — because your current card died, because you're building a PC for work or study, or because you want to play current games at good settings — buy from what's actually on shelves now. As of October 2026, reporting puts that choice as the RTX 50 series and AMD's RX 9000 series. Don't wait for an RTX 60 that leaks now place in 2028, and don't pay a panic premium for a rumoured card that may never launch.

If you can wait, and your current card still plays what you want, waiting is also rational. The earliest realistic new-architecture target from leaks is late 2027, with 2028 more likely. That's a long wait, and it's a wait with no guarantee. Set a date — say, "if nothing is announced by this time next year, I'll buy then" — rather than waiting indefinitely for a perfect launch that keeps slipping.

A few practical checks before you buy in this market:

• Check actual street prices, not launch MSRPs. Memory constraints and demand have pushed some high-end cards well above their launch prices in third-party trackers, so the "best value" card on paper may not be the best value in a shop today.
• Check VRAM for your use case. If you play at 1440p or 4K, or you do video editing or local AI work, VRAM matters more than a small clock-speed difference.
• Check DLSS and driver support for the card you're considering, especially if you're buying an older generation to save money.
• Check power supply and case fit. High-end cards are large and power-hungry, and a card you can't cool or power properly is a bad buy at any price.

What to watch next

Three things will tell you whether 2026 was a one-year pause or the start of a longer shift.

First, memory supply. If memory availability improves, gaming production and launch plans should ease with it. If it doesn't, expect more delays and higher prices, regardless of what Nvidia announces.

Second, any official Nvidia statement on RTX 60 or a GeForce refresh. Until Nvidia itself announces a date, architecture, or product, everything else is leak-based reporting. When Nvidia does speak, that will reset the timeline — in either direction.

Third, Rubin CPX deployments. If CPX ships at the end of 2026 as expected, and cloud providers offer it at scale in early 2027, it will confirm Nvidia's pivot to inference-specialised silicon. If it slips, it will show that even AI hardware is subject to the same supply and engineering constraints as gaming hardware.

For now, the story of Nvidia in 2026 is a story of two companies in one: an AI infrastructure giant moving fast on Rubin and CPX, and a gaming GPU business standing still, waiting for memory, capacity, and a roadmap slot. Gamers didn't get a new card this year. AI developers got a new category of chip. And until the supply picture changes, that's the trade-off Nvidia has chosen to make.