Here’s Why This Year’s CES Proved AI Has Changed Everything: A Look at the Biggest AI Announcements at CES 2026

This year's CES marked a major shift in momentum, pivoting away from consumer announcements and more toward AI. This transition disappointed several gamers, who were expecting CES 2026 to be jam-packed with products for them. However, given the mainstream adoption of AI, companies like NVIDIA found it difficult to ignore the technology at CES.

The showcases on the AI front by the likes of NVIDIA, AMD, and many other players have dictated that AI is now moving into a new front, from training to applications in the form of physical and agentic AI. And, of course, for this pivot, the world needs to have the necessary computing capacity onboard, which is why, out of the major announcements from NVIDIA/AMD, the focus on the next era of AI infrastructure was dominant across keynotes by Jensen Huang and Lisa Su.

Here are some of our top picks from this year's CES 2026, featuring AI, and we will categorize them into multiple tiers, which you'll see throughout our coverage.

The Best AI Platform: NVIDIA's Vera Rubin

NVIDIA's Vera Rubin AI lineup was one of the most anticipated releases of this year, following its initial showcase at GTC 2025, where Jensen revealed that the entire hardware stack would be innovated, from networking to chips. At this year's CES 2026 keynote, NVIDIA's CEO gave us a deep dive into its upcoming Rubin products, including the six different new chips, which are as follows:

• Rubin GPU (with 336 Billion Transistors)
• Vera CPU (with 227 Billion Transistors)
• NVLINK 6 Switch for Interconnect
• CX9 & BF4 for Networking
• Spectrum-X 102.4T CPO for silicon photonics

The reason why Vera Rubin wins the best AI platform award from us is not only due to the company's advancements, but also how NVIDIA has managed to bring the architecture into "full production" in just a nine-month product cadence, and that includes tape-outs and validation stages as well. For reference, NVIDIA's Blackwell Ultra entered mass production in Q3 2025, with customer shipments commencing in Q4 2025. This demonstrates the current level of disruption NVIDIA is causing in the AI industry.

We already have a technical rundown of Vera Rubin, which is available here, discussing the Rubin GPU, Vera CPUs, and how they are configured together in the NVL72 rack. But, to give an idea of the power of Rubin chips, NVIDIA's CES keynote dived into performance figures, where Jensen revealed that Rubin delivers a massive 5x uplift versus Blackwell in NVFP4 inference, with factory throughput increasing by up to 10 times compared to GB200, especially in MoE environments.

NVIDIA's Vera Rubin AI platform showcases the progress of the AI infrastructure world as we move into 2026. It is reported that customer shipments will begin rolling out in H2, indicating that hyperscalers could start training frontier models on Rubin by the end of this year, which is simply shocking.

The Best AI Rack: AMD's Helios

When we discuss AI racks, it's not just about which manufacturer delivers the best performance onboard, but rather the innovations introduced in compute rack arrangements, networking layouts, and other related aspects. AMD's Helios rack was unveiled to the public at CES 2026, and one of the major standout points about it is that the rack is built upon Meta's Open Rack Wide (ORW) specification. We will discuss next why this is important.

• AI Compute Performance: 2.9 Exaflops
• Memory: 31 TB HBM4 Memory
• Scale Out Bandwidth: 43 TB/s
• Manufacturing Process: 2nm / 3nm Advanced Process
• CPU Cores: 4,600 "Zen 6" CPU Cores

AMD's Helios focuses on having a 21-inch internal width for its compute racks, allowing the company to fit its wider Instinct MI455X trays horizontally with less impedance. More importantly, the ORW focuses on maximizing data center floor space, and for that, AMD's Helios rack integrates a "double-wide" physical frame, which also allows for a laminar airflow path. Another interesting feature of AMD's Helios is the presence of a centralized 48V DC busbar running vertically along the back of the rack, which powers individual compute nodes.

However, the biggest benefit of Helios, built upon Meta's ORW, is that it makes the rack modular enough for hyperscalers to integrate it into existing data center configurations. With the use of UALink, along with ORW-compliant designs, AMD aims to ensure that companies like Microsoft and Meta don't find it challenging to integrate Helios racks alongside other 'running systems', which enables the company to address the reluctance of hyperscalers when it comes to switching to newer racks.

Unlike NVIDIA, whose entire AI rack functions as a single unit, AMD's Helios would facilitate integration more easily. Combined with ORW design capabilities in terms of power and cooling, the Helios will undoubtedly be a more attractive option relative to past racks from Team Red. Coupled with the next-generation Instinct MI400 series AI chips, AMD is poised to increase the competition in the AI infrastructure race, giving NVIDIA a run for its money.

The Best AI Software Showcase: NVIDIA's Alpamayo

NVIDIA's primary focus at this year's CES was to discuss how the company is the largest contributor to the open-source model race, and it has ultimately managed to surpass Chinese AI giants like DeepSeek and Baidu. The company's Nemotron offerings have dominated the open-source space, and at CES 2026, NVIDIA announced that they plan to bring their expertise with open models into autonomous vehicles, as Team Green showcased "Alpamayo", and here's how they describe it:

NVIDIA Alpamayo is an open portfolio of AI models, simulation frameworks, and physical AI datasets designed to accelerate the development of safe, transparent, and reasoning-based autonomous vehicles. Built for Level 4 autonomy, Alpamayo lets vehicles perceive, reason, and act with human-like judgment, while providing the interpretability and openness required for safety validation and regulatory collaboration

NVIDIA's goal with Alpamayo is to make vehicle autonomy both open and powerful, driven by the fact that Team Green provides "fully open models, simulation frameworks, and datasets" for developers, which they can then fine-tune for their respective frameworks and regulations. More importantly, NVIDIA claims that Alpamayo focuses on "reasoning-based" autonomy, which has provided the company an accelerated path towards Level 4 driving, which is a feat limited to the likes of Waymo and Mercedes-Benz.

The Alpamayo showcase clearly indicates that NVIDIA plans to revolutionize the entire AI ecosystem by embedding the technology into aspects that can be automated. With physical and agentic frameworks already seeing massive development by NVIDIA, Alpamayo has opened up a new front for the company in the realm of self-driving vehicles. Most importantly, NVIDIA has open-sourced the technology for developers, making its integration easier and more widespread.

The Best Edge AI Platform: AMD's Ryzen AI Halo

AMD surprised us by unveiling the Ryzen AI Halo, an edge AI device designed as an alternative to NVIDIA's DGX Spark platform. AMD says that these devices will include custom 'Ryzen AI Halo' chips, and while the firm didn't go deep into the specifics, it did disclose that the SoCs will feature Zen 5 CPU, RDNA 3.5 GPU, and XDNA 2 NPU architectures. The idea behind Ryzen AI Halo is to integrate ROCm into developer workloads, providing them with a capable hardware solution that enables localized AI computation.

NVIDIA's DGX Spark was a significant release at Computex 2025, as the intention behind the product was to make compute accessible to everyone, rather than making it exclusive to a few hyperscalers. AMD appears to be building upon this ideology as well, with Ryzen AI Halo. Team Red plans to feature full AMD ROCm Support, including the newly released ROCm 7.2.2 suite, which will be optimized for Dev-Ready applications such as LM Studio, ComfyUI, VS Code, and more. It will enable optimizations for several models, including GPT-OSS and FLUX.2, SDXL, and More, and finally, it will carry Day 0 support for leading AI models.

The biggest concern with such devices is their price points, considering that the DGX Spark has been retailing for $4,000 for its high-end configuration, making it out of reach for an average developer. While AMD didn't reveal Ryzen AI Halo pricing, it's safe to say that the price tag will be a decisive factor. Considering that AMD intends to integrate an SoC similar to the Ryzen AI MAX lineup, the pricing will surely be much lower than what the DGX Spark costs.

Conclusion: AI, AI & AI at CES 2026

Despite being the "Consumer Electronics" show, this year's CES clearly demonstrated that AI is a technology too big to ignore for now, and it has also made consumer offerings significantly more challenging to afford, further exacerbated by ongoing memory shortages. However, when discussing how AI has advanced over the years, CES 2026 was a significant indicator that there's a lot ahead for the technology, especially when considering emerging segments like physical and agentic AI.

Follow Wccftech on Google to get more of our news coverage in your feeds.