CES 2026 will finally answer big questions around Nvidia’s RTX 50 Super GPUs

With its Blackwell architecture well established on AI data centers, cloud services, workstations, and desktop/laptop PCs, Nvidia’s CES 2026 press event is likely to focus less on new launches. Rather, it is expected that the company will delve into refinements, roadmap signals, and how it plans to push its hardware and software stack forward over the coming year.

Nvidia has confirmed that founder and CEO Jensen Huang will be delivering the company’s CES opening keynote on January 5, a day ahead of the main CES show floor opening. CES has become one of Nvidia’s most important stages of the year, even if the company isn’t launching a new generation of GPUs.

As in previous years, this is where Nvidia is expected to outline its priorities for gaming, AI, data centers, and emerging platforms. Beyond the keynote, Nvidia will also maintain a strong on-ground presence throughout CES with potential demos and showcases. 

Here’s a quick look into what Nvidia may have in store for us next month.

RTX 50 Super

If Nvidia has any consumer GPU news at CES 2026, it will likely be about the RTX 50 Super. As per leaks and reports, the Super refresh was originally planned for a late 2025 launch to align with the holiday season. 

However, with rising memory prices and Nvidia’s own history of spacing out refreshes, a delay into early or mid-2026 now looks more likely, with CES serving as a showcase rather than a launch event.

The RTX 50 Super is not expected to introduce any major changes. Instead, it appears to be a small update to the existing Blackwell GPUs, with adjusted specifications aimed at improving how the lineup is balanced. The RTX 40 Super refresh followed a similar pattern, and current leaks suggest Nvidia is taking the same approach again.

Based on what’s surfaced so far, Nvidia is expected to focus on select SKUs rather than a full-stack refresh. Rumored models include the RTX 5070 Super, RTX 5070 Ti Super, and RTX 5080 Super. The RTX 5070 Super is expected to move to a higher VRAM configuration, potentially 18GB, addressing one of the biggest complaints with the standard model. 

The RTX 5080 Super is rumored to keep a similar core layout but could ship with 8GB of additional memory for a total of 24GB along with higher bandwidth, possibly pushing it closer to the RTX 5090 in real-world workloads rather than raw compute.

Any performance gains are expected to be modest. Small clock bumps and minor core adjustments are more likely than major increases in CUDA count. The goal appears to be smoothing out gaps within the lineup rather than delivering a clear generational jump.

Memory configuration is one of the main reasons a refresh exists at all. As games and software continue to use more memory, some RTX 50 models are already under pressure. A Super refresh would give Nvidia room to adjust these configurations without changing the underlying architecture

The timing, however, remains uncertain. Higher memory costs make it harder to change configurations without pushing prices up, giving Nvidia a clear incentive to wait. As a result, CES 2026 is more likely to bring confirmation rather than availability. 

Nvidia may acknowledge RTX 50 Super or outline its plans, while keeping the actual release for later in 2026. For buyers, that clarity alone could be enough to decide whether to upgrade now or wait a little longer.

AI to take center stage

As much as gamers care about GPUs, AI will remain Nvidia’s main focus at CES 2026. Expect the company to spend significant time talking about AI hardware for enterprise and data centers, particularly accelerators designed for inference, training, and edge deployment. 

CES isn’t GTC, but it has increasingly become a place where Nvidia outlines how its AI stack scales, from massive data centers down to local AI PCs.

Automotive is another likely focus. Nvidia has been deeply embedded in autonomous driving and in-car AI platforms for years, and CES is traditionally where carmakers and chipmakers align their roadmaps. New announcements around DRIVE platforms, in-vehicle AI compute, or expanded partnerships wouldn’t be surprising.

Beyond cars, Nvidia may also touch on embedded systems, industrial AI, robotics, and edge computing. These are areas where Nvidia’s hardware and software platforms already dominate, and CES provides the right stage to show how that technology trickles down into everyday products.

The Arm wild card

One of the most intriguing rumors heading into CES 2026 surrounds Nvidia’s first serious Arm-based AI PC SoC, codenamed N1x. While Team Green has remained silent, reports from earlier this year suggest this chip could be far more ambitious than initially assumed.

According to a particular leak, the N1x may feature the same GPU core count as the RTX 5070, but integrated directly into an SoC rather than existing as a discrete GPU. If true, that would put the N1x in uncharted territory for integrated graphics, with performance reportedly exceeding every other iGPU currently on the market by a wide margin.

The implication here is significant. Rather than positioning N1x as a low-power ARM experiment, Nvidia could be building a high-performance AI PC chip designed to handle gaming, content creation, and local AI workloads without a discrete GPU. That aligns neatly with Nvidia’s broader push toward AI-first computing, where neural workloads, inference, and acceleration matter as much as traditional raster performance.

While CES would be the ideal venue for Nvidia to tease such a platform, don’t expect a full product launch or retail-ready devices right away. A conceptual reveal framed around “AI PCs,” next-gen Windows on ARM systems, or premium embedded platforms, feels far more likely. Nvidia could emphasize efficiency, AI throughput, and graphics leadership, without directly positioning N1x as an x86 replacement.

If even part of the rumor proves accurate, N1x wouldn’t just challenge existing ARM laptop chips, primarily from Qualcomm, it could blur the line between integrated and discrete graphics entirely.